Upload folder using huggingface_hub
Browse files- .gitattributes +1 -0
- README.md +35 -3
- assets/CVPR_2025_SimCIS.pdf +3 -0
- assets/framework.svg +0 -0
- ps_100-10_42/step6/config.yaml +443 -0
- ps_100-10_42/step6/fake_query.pkl +3 -0
- ps_100-10_42/step6/last_checkpoint +1 -0
- ps_100-10_42/step6/metrics.json +0 -0
- ps_100-10_42/step6/model_final.pth +3 -0
- ps_100-10_42/step6/psd_distribution.json +1 -0
- ps_100-50_40/step2/config.yaml +439 -0
- ps_100-50_40/step2/last_checkpoint +1 -0
- ps_100-50_40/step2/metrics.json +0 -0
- ps_100-50_40/step2/model_final.pth +3 -0
- ps_100-50_40/step2/psd_distribution.json +1 -0
- ps_100-5_35/step11/config.yaml +452 -0
- ps_100-5_35/step11/fake_query.pkl +3 -0
- ps_100-5_35/step11/last_checkpoint +1 -0
- ps_100-5_35/step11/metrics.json +0 -0
- ps_100-5_35/step11/model_final.pth +3 -0
- ps_100-5_35/step11/psd_distribution.json +1 -0
- ss_100-10_42/step6/config.yaml +452 -0
- ss_100-10_42/step6/fake_query.pkl +3 -0
- ss_100-10_42/step6/last_checkpoint +1 -0
- ss_100-10_42/step6/metrics.json +0 -0
- ss_100-10_42/step6/model_final.pth +3 -0
- ss_100-10_42/step6/psd_distribution.json +1 -0
- ss_100-50_48/step2/config.yaml +452 -0
- ss_100-50_48/step2/fake_query.pkl +3 -0
- ss_100-50_48/step2/last_checkpoint +1 -0
- ss_100-50_48/step2/metrics.json +0 -0
- ss_100-50_48/step2/model_final.pth +3 -0
- ss_100-50_48/step2/psd_distribution.json +1 -0
- ss_100-5_38/step11/config.yaml +452 -0
- ss_100-5_38/step11/fake_query.pkl +3 -0
- ss_100-5_38/step11/last_checkpoint +1 -0
- ss_100-5_38/step11/metrics.json +0 -0
- ss_100-5_38/step11/model_final.pth +3 -0
- ss_100-5_38/step11/psd_distribution.json +1 -0
.gitattributes
CHANGED
@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
ss_100base/step1/metrics.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
ss_100base/step1/metrics.json filter=lfs diff=lfs merge=lfs -text
|
37 |
+
assets/CVPR_2025_SimCIS.pdf filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
@@ -1,3 +1,35 @@
|
|
1 |
-
---
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
license: mit
|
5 |
+
---
|
6 |
+
|
7 |
+
# SimCIS
|
8 |
+
|
9 |
+
**Rethinking Query-based Transformer for Continual Image Segmentation.** *(CVPR2025)*
|
10 |
+
|
11 |
+
[](LICENSE)[](./assets/CVPR_2025_SimCIS.pdf)
|
12 |
+

|
13 |
+
|
14 |
+
By [Yuchen Zhu*](https://github.com/ZhuYuChenNO1), [Cheng Shi*](https://chengshiest.github.io/), Dingyou Wang, Jiajin Tang, Zhengxuan Wei, Yu Wu, Guanbin Li and [Sibei Yang†](https://faculty.sist.shanghaitech.edu.cn/yangsibei/)
|
15 |
+
|
16 |
+
## SEE [Github](https://github.com/SooLab/SimCIS) to use the weights!!!
|
17 |
+
|
18 |
+
## 📖 Cite Us
|
19 |
+
If you find this repository useful in your research, please consider giving a star ⭐ and a citation
|
20 |
+
```bibtex
|
21 |
+
@inproceedings{zhu2025rethinking,
|
22 |
+
title={Rethinking Query-based Transformer for Continual Image Segmentation},
|
23 |
+
author={Zhu, Yuchen and Shi, Cheng and Wang, Dingyou and Tang, Jiajin and Wei, Zhengxuan and Wu, Yu and Li, Guanbin and Yang, Sibei},
|
24 |
+
booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
|
25 |
+
pages={4595--4606},
|
26 |
+
year={2025}
|
27 |
+
}
|
28 |
+
```
|
29 |
+
|
30 |
+
|
31 |
+
## 👍 Acknowledgement and Related Work
|
32 |
+
* This code is mainly based on [Mask2Former](https://github.com/facebookresearch/Mask2Former). We thank them for their excellent work.
|
33 |
+
* Related work for continual image segmentation: [Balconpas](https://github.com/jinpeng0528/BalConpas/tree/master), [ECLIPSE](https://github.com/clovaai/ECLIPSE). We appreciate the contributions of these researchers.
|
34 |
+
|
35 |
+
|
assets/CVPR_2025_SimCIS.pdf
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c24759accc65544bfa402b3066ed5be382637b121fa020134ffb9a1d4e7a11e1
|
3 |
+
size 1575052
|
assets/framework.svg
ADDED
|
ps_100-10_42/step6/config.yaml
ADDED
@@ -0,0 +1,443 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
CONT:
|
2 |
+
BASE_CLS: 100
|
3 |
+
COLLECT_QUERY_MODE: false
|
4 |
+
CUMULATIVE_PSDNUM: false
|
5 |
+
DISTRIBUTION_ALPHA: 0.5
|
6 |
+
FREEZE_LABEL: false
|
7 |
+
INC_CLS: 10
|
8 |
+
KL_ALL: true
|
9 |
+
KL_WEIGHT: 2.0
|
10 |
+
LIB_SIZE: 80
|
11 |
+
MED_TOKENS_WEIGHT: 5.0
|
12 |
+
MEMORY: false
|
13 |
+
OLD_WEIGHTS: ./output/ps/100-10_fake1/step5/model_final.pth
|
14 |
+
PSD_LABEL_THRESHOLD: 0.35
|
15 |
+
SETTING: overlapped
|
16 |
+
TASK: 6
|
17 |
+
TOT_CLS: 150
|
18 |
+
VQ_NUMBER: 1
|
19 |
+
WEIGHTED_SAMPLE: true
|
20 |
+
WEIGHTS: null
|
21 |
+
CUDNN_BENCHMARK: false
|
22 |
+
DATALOADER:
|
23 |
+
ASPECT_RATIO_GROUPING: true
|
24 |
+
FILTER_EMPTY_ANNOTATIONS: true
|
25 |
+
NUM_WORKERS: 2
|
26 |
+
REPEAT_SQRT: true
|
27 |
+
REPEAT_THRESHOLD: 0.0
|
28 |
+
SAMPLER_TRAIN: TrainingSampler
|
29 |
+
DATASETS:
|
30 |
+
PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
|
31 |
+
PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
|
32 |
+
PROPOSAL_FILES_TEST: []
|
33 |
+
PROPOSAL_FILES_TRAIN: []
|
34 |
+
TEST:
|
35 |
+
- current_ade20k_panoptic_val
|
36 |
+
TRAIN:
|
37 |
+
- current_ade20k_panoptic_train
|
38 |
+
GLOBAL:
|
39 |
+
HACK: 1.0
|
40 |
+
INPUT:
|
41 |
+
COLOR_AUG_SSD: true
|
42 |
+
CROP:
|
43 |
+
ENABLED: true
|
44 |
+
SINGLE_CATEGORY_MAX_AREA: 1.0
|
45 |
+
SIZE:
|
46 |
+
- 640
|
47 |
+
- 640
|
48 |
+
TYPE: absolute
|
49 |
+
DATASET_MAPPER_NAME: continual_panoptic
|
50 |
+
FORMAT: RGB
|
51 |
+
IMAGE_SIZE: 1024
|
52 |
+
MASK_FORMAT: polygon
|
53 |
+
MAX_SCALE: 2.0
|
54 |
+
MAX_SIZE_TEST: 2560
|
55 |
+
MAX_SIZE_TRAIN: 2560
|
56 |
+
MIN_SCALE: 0.1
|
57 |
+
MIN_SIZE_TEST: 640
|
58 |
+
MIN_SIZE_TRAIN:
|
59 |
+
- 320
|
60 |
+
- 384
|
61 |
+
- 448
|
62 |
+
- 512
|
63 |
+
- 576
|
64 |
+
- 640
|
65 |
+
- 704
|
66 |
+
- 768
|
67 |
+
- 832
|
68 |
+
- 896
|
69 |
+
- 960
|
70 |
+
- 1024
|
71 |
+
- 1088
|
72 |
+
- 1152
|
73 |
+
- 1216
|
74 |
+
- 1280
|
75 |
+
MIN_SIZE_TRAIN_SAMPLING: choice
|
76 |
+
RANDOM_FLIP: horizontal
|
77 |
+
SIZE_DIVISIBILITY: 640
|
78 |
+
MODEL:
|
79 |
+
ANCHOR_GENERATOR:
|
80 |
+
ANGLES:
|
81 |
+
- - -90
|
82 |
+
- 0
|
83 |
+
- 90
|
84 |
+
ASPECT_RATIOS:
|
85 |
+
- - 0.5
|
86 |
+
- 1.0
|
87 |
+
- 2.0
|
88 |
+
NAME: DefaultAnchorGenerator
|
89 |
+
OFFSET: 0.0
|
90 |
+
SIZES:
|
91 |
+
- - 32
|
92 |
+
- 64
|
93 |
+
- 128
|
94 |
+
- 256
|
95 |
+
- 512
|
96 |
+
BACKBONE:
|
97 |
+
FREEZE_AT: 0
|
98 |
+
NAME: build_resnet_backbone
|
99 |
+
DEVICE: cuda
|
100 |
+
FPN:
|
101 |
+
FUSE_TYPE: sum
|
102 |
+
IN_FEATURES: []
|
103 |
+
NORM: ''
|
104 |
+
OUT_CHANNELS: 256
|
105 |
+
KEYPOINT_ON: false
|
106 |
+
LOAD_PROPOSALS: false
|
107 |
+
MASK_FORMER:
|
108 |
+
CLASS_WEIGHT: 4.0
|
109 |
+
CLIP_DIM: 768
|
110 |
+
DEC_LAYERS: 10
|
111 |
+
DEEP_SUPERVISION: true
|
112 |
+
DICE_WEIGHT: 5.0
|
113 |
+
DIM_FEEDFORWARD: 2048
|
114 |
+
DROPOUT: 0.0
|
115 |
+
ENC_LAYERS: 0
|
116 |
+
ENFORCE_INPUT_PROJ: false
|
117 |
+
HIDDEN_DIM: 256
|
118 |
+
IMPORTANCE_SAMPLE_RATIO: 0.75
|
119 |
+
MASK_WEIGHT: 5.0
|
120 |
+
NHEADS: 8
|
121 |
+
NO_OBJECT_WEIGHT: 0.1
|
122 |
+
NUM_OBJECT_QUERIES: 100
|
123 |
+
OVERSAMPLE_RATIO: 3.0
|
124 |
+
PRE_NORM: false
|
125 |
+
SIZE_DIVISIBILITY: 32
|
126 |
+
TEST:
|
127 |
+
INSTANCE_ON: true
|
128 |
+
OBJECT_MASK_THRESHOLD: 0.25
|
129 |
+
OVERLAP_THRESHOLD: 0.8
|
130 |
+
PANOPTIC_ON: true
|
131 |
+
SEMANTIC_ON: true
|
132 |
+
SEM_SEG_POSTPROCESSING_BEFORE_INFERENCE: false
|
133 |
+
TEXT_PATH: label_embedding/ade20k_150_laion768_all.npy
|
134 |
+
TRAIN_NUM_POINTS: 12544
|
135 |
+
TRANSFORMER_DECODER_NAME: MultiScaleMaskedTransformerDecoder
|
136 |
+
TRANSFORMER_IN_FEATURE: multi_scale_pixel_decoder
|
137 |
+
USE_TEXT_EMBEDDING: false
|
138 |
+
MASK_ON: false
|
139 |
+
META_ARCHITECTURE: MaskFormer
|
140 |
+
PANOPTIC_FPN:
|
141 |
+
COMBINE:
|
142 |
+
ENABLED: true
|
143 |
+
INSTANCES_CONFIDENCE_THRESH: 0.5
|
144 |
+
OVERLAP_THRESH: 0.5
|
145 |
+
STUFF_AREA_LIMIT: 4096
|
146 |
+
INSTANCE_LOSS_WEIGHT: 1.0
|
147 |
+
PIXEL_MEAN:
|
148 |
+
- 123.675
|
149 |
+
- 116.28
|
150 |
+
- 103.53
|
151 |
+
PIXEL_STD:
|
152 |
+
- 58.395
|
153 |
+
- 57.12
|
154 |
+
- 57.375
|
155 |
+
PROPOSAL_GENERATOR:
|
156 |
+
MIN_SIZE: 0
|
157 |
+
NAME: RPN
|
158 |
+
RESNETS:
|
159 |
+
DEFORM_MODULATED: false
|
160 |
+
DEFORM_NUM_GROUPS: 1
|
161 |
+
DEFORM_ON_PER_STAGE:
|
162 |
+
- false
|
163 |
+
- false
|
164 |
+
- false
|
165 |
+
- false
|
166 |
+
DEPTH: 50
|
167 |
+
NORM: FrozenBN
|
168 |
+
NUM_GROUPS: 1
|
169 |
+
OUT_FEATURES:
|
170 |
+
- res2
|
171 |
+
- res3
|
172 |
+
- res4
|
173 |
+
- res5
|
174 |
+
RES2_OUT_CHANNELS: 256
|
175 |
+
RES4_DILATION: 1
|
176 |
+
RES5_DILATION: 1
|
177 |
+
RES5_MULTI_GRID:
|
178 |
+
- 1
|
179 |
+
- 1
|
180 |
+
- 1
|
181 |
+
STEM_OUT_CHANNELS: 64
|
182 |
+
STEM_TYPE: basic
|
183 |
+
STRIDE_IN_1X1: false
|
184 |
+
WIDTH_PER_GROUP: 64
|
185 |
+
RETINANET:
|
186 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
187 |
+
BBOX_REG_WEIGHTS: &id002
|
188 |
+
- 1.0
|
189 |
+
- 1.0
|
190 |
+
- 1.0
|
191 |
+
- 1.0
|
192 |
+
FOCAL_LOSS_ALPHA: 0.25
|
193 |
+
FOCAL_LOSS_GAMMA: 2.0
|
194 |
+
IN_FEATURES:
|
195 |
+
- p3
|
196 |
+
- p4
|
197 |
+
- p5
|
198 |
+
- p6
|
199 |
+
- p7
|
200 |
+
IOU_LABELS:
|
201 |
+
- 0
|
202 |
+
- -1
|
203 |
+
- 1
|
204 |
+
IOU_THRESHOLDS:
|
205 |
+
- 0.4
|
206 |
+
- 0.5
|
207 |
+
NMS_THRESH_TEST: 0.5
|
208 |
+
NORM: ''
|
209 |
+
NUM_CLASSES: 80
|
210 |
+
NUM_CONVS: 4
|
211 |
+
PRIOR_PROB: 0.01
|
212 |
+
SCORE_THRESH_TEST: 0.05
|
213 |
+
SMOOTH_L1_LOSS_BETA: 0.1
|
214 |
+
TOPK_CANDIDATES_TEST: 1000
|
215 |
+
ROI_BOX_CASCADE_HEAD:
|
216 |
+
BBOX_REG_WEIGHTS:
|
217 |
+
- &id001
|
218 |
+
- 10.0
|
219 |
+
- 10.0
|
220 |
+
- 5.0
|
221 |
+
- 5.0
|
222 |
+
- - 20.0
|
223 |
+
- 20.0
|
224 |
+
- 10.0
|
225 |
+
- 10.0
|
226 |
+
- - 30.0
|
227 |
+
- 30.0
|
228 |
+
- 15.0
|
229 |
+
- 15.0
|
230 |
+
IOUS:
|
231 |
+
- 0.5
|
232 |
+
- 0.6
|
233 |
+
- 0.7
|
234 |
+
ROI_BOX_HEAD:
|
235 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
236 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
237 |
+
BBOX_REG_WEIGHTS: *id001
|
238 |
+
CLS_AGNOSTIC_BBOX_REG: false
|
239 |
+
CONV_DIM: 256
|
240 |
+
FC_DIM: 1024
|
241 |
+
FED_LOSS_FREQ_WEIGHT_POWER: 0.5
|
242 |
+
FED_LOSS_NUM_CLASSES: 50
|
243 |
+
NAME: ''
|
244 |
+
NORM: ''
|
245 |
+
NUM_CONV: 0
|
246 |
+
NUM_FC: 0
|
247 |
+
POOLER_RESOLUTION: 14
|
248 |
+
POOLER_SAMPLING_RATIO: 0
|
249 |
+
POOLER_TYPE: ROIAlignV2
|
250 |
+
SMOOTH_L1_BETA: 0.0
|
251 |
+
TRAIN_ON_PRED_BOXES: false
|
252 |
+
USE_FED_LOSS: false
|
253 |
+
USE_SIGMOID_CE: false
|
254 |
+
ROI_HEADS:
|
255 |
+
BATCH_SIZE_PER_IMAGE: 512
|
256 |
+
IN_FEATURES:
|
257 |
+
- res4
|
258 |
+
IOU_LABELS:
|
259 |
+
- 0
|
260 |
+
- 1
|
261 |
+
IOU_THRESHOLDS:
|
262 |
+
- 0.5
|
263 |
+
NAME: Res5ROIHeads
|
264 |
+
NMS_THRESH_TEST: 0.5
|
265 |
+
NUM_CLASSES: 80
|
266 |
+
POSITIVE_FRACTION: 0.25
|
267 |
+
PROPOSAL_APPEND_GT: true
|
268 |
+
SCORE_THRESH_TEST: 0.05
|
269 |
+
ROI_KEYPOINT_HEAD:
|
270 |
+
CONV_DIMS:
|
271 |
+
- 512
|
272 |
+
- 512
|
273 |
+
- 512
|
274 |
+
- 512
|
275 |
+
- 512
|
276 |
+
- 512
|
277 |
+
- 512
|
278 |
+
- 512
|
279 |
+
LOSS_WEIGHT: 1.0
|
280 |
+
MIN_KEYPOINTS_PER_IMAGE: 1
|
281 |
+
NAME: KRCNNConvDeconvUpsampleHead
|
282 |
+
NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
|
283 |
+
NUM_KEYPOINTS: 17
|
284 |
+
POOLER_RESOLUTION: 14
|
285 |
+
POOLER_SAMPLING_RATIO: 0
|
286 |
+
POOLER_TYPE: ROIAlignV2
|
287 |
+
ROI_MASK_HEAD:
|
288 |
+
CLS_AGNOSTIC_MASK: false
|
289 |
+
CONV_DIM: 256
|
290 |
+
NAME: MaskRCNNConvUpsampleHead
|
291 |
+
NORM: ''
|
292 |
+
NUM_CONV: 0
|
293 |
+
POOLER_RESOLUTION: 14
|
294 |
+
POOLER_SAMPLING_RATIO: 0
|
295 |
+
POOLER_TYPE: ROIAlignV2
|
296 |
+
RPN:
|
297 |
+
BATCH_SIZE_PER_IMAGE: 256
|
298 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
299 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
300 |
+
BBOX_REG_WEIGHTS: *id002
|
301 |
+
BOUNDARY_THRESH: -1
|
302 |
+
CONV_DIMS:
|
303 |
+
- -1
|
304 |
+
HEAD_NAME: StandardRPNHead
|
305 |
+
IN_FEATURES:
|
306 |
+
- res4
|
307 |
+
IOU_LABELS:
|
308 |
+
- 0
|
309 |
+
- -1
|
310 |
+
- 1
|
311 |
+
IOU_THRESHOLDS:
|
312 |
+
- 0.3
|
313 |
+
- 0.7
|
314 |
+
LOSS_WEIGHT: 1.0
|
315 |
+
NMS_THRESH: 0.7
|
316 |
+
POSITIVE_FRACTION: 0.5
|
317 |
+
POST_NMS_TOPK_TEST: 1000
|
318 |
+
POST_NMS_TOPK_TRAIN: 2000
|
319 |
+
PRE_NMS_TOPK_TEST: 6000
|
320 |
+
PRE_NMS_TOPK_TRAIN: 12000
|
321 |
+
SMOOTH_L1_BETA: 0.0
|
322 |
+
SEM_SEG_HEAD:
|
323 |
+
ASPP_CHANNELS: 256
|
324 |
+
ASPP_DILATIONS:
|
325 |
+
- 6
|
326 |
+
- 12
|
327 |
+
- 18
|
328 |
+
ASPP_DROPOUT: 0.1
|
329 |
+
COMMON_STRIDE: 4
|
330 |
+
CONVS_DIM: 256
|
331 |
+
DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES:
|
332 |
+
- res3
|
333 |
+
- res4
|
334 |
+
- res5
|
335 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_HEADS: 8
|
336 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_POINTS: 4
|
337 |
+
IGNORE_VALUE: 255
|
338 |
+
IN_FEATURES:
|
339 |
+
- res2
|
340 |
+
- res3
|
341 |
+
- res4
|
342 |
+
- res5
|
343 |
+
LOSS_TYPE: hard_pixel_mining
|
344 |
+
LOSS_WEIGHT: 1.0
|
345 |
+
MASK_DIM: 256
|
346 |
+
NAME: MaskFormerHead
|
347 |
+
NORM: GN
|
348 |
+
NUM_CLASSES: 150
|
349 |
+
PIXEL_DECODER_NAME: MSDeformAttnPixelDecoder
|
350 |
+
PROJECT_CHANNELS:
|
351 |
+
- 48
|
352 |
+
PROJECT_FEATURES:
|
353 |
+
- res2
|
354 |
+
TRANSFORMER_ENC_LAYERS: 6
|
355 |
+
USE_DEPTHWISE_SEPARABLE_CONV: false
|
356 |
+
SWIN:
|
357 |
+
APE: false
|
358 |
+
ATTN_DROP_RATE: 0.0
|
359 |
+
DEPTHS:
|
360 |
+
- 2
|
361 |
+
- 2
|
362 |
+
- 6
|
363 |
+
- 2
|
364 |
+
DROP_PATH_RATE: 0.3
|
365 |
+
DROP_RATE: 0.0
|
366 |
+
EMBED_DIM: 96
|
367 |
+
MLP_RATIO: 4.0
|
368 |
+
NUM_HEADS:
|
369 |
+
- 3
|
370 |
+
- 6
|
371 |
+
- 12
|
372 |
+
- 24
|
373 |
+
OUT_FEATURES:
|
374 |
+
- res2
|
375 |
+
- res3
|
376 |
+
- res4
|
377 |
+
- res5
|
378 |
+
PATCH_NORM: true
|
379 |
+
PATCH_SIZE: 4
|
380 |
+
PRETRAIN_IMG_SIZE: 224
|
381 |
+
QKV_BIAS: true
|
382 |
+
QK_SCALE: null
|
383 |
+
USE_CHECKPOINT: false
|
384 |
+
WINDOW_SIZE: 7
|
385 |
+
WEIGHTS: ./output/ps/100-10_fake1/step5/model_final.pth
|
386 |
+
OUTPUT_DIR: ./output/ps/100-10_fake1/step6
|
387 |
+
SEED: -1
|
388 |
+
SOLVER:
|
389 |
+
AMP:
|
390 |
+
ENABLED: true
|
391 |
+
BACKBONE_MULTIPLIER: 0.1
|
392 |
+
BASE_LR: 5.0e-05
|
393 |
+
BASE_LR_END: 0.0
|
394 |
+
BIAS_LR_FACTOR: 1.0
|
395 |
+
CHECKPOINT_PERIOD: 10000
|
396 |
+
CLIP_GRADIENTS:
|
397 |
+
CLIP_TYPE: full_model
|
398 |
+
CLIP_VALUE: 0.01
|
399 |
+
ENABLED: true
|
400 |
+
NORM_TYPE: 2.0
|
401 |
+
GAMMA: 0.1
|
402 |
+
IMS_PER_BATCH: 8
|
403 |
+
LR_SCHEDULER_NAME: WarmupMultiStepLR
|
404 |
+
MAX_ITER: 10000
|
405 |
+
MOMENTUM: 0.9
|
406 |
+
NESTEROV: false
|
407 |
+
NUM_DECAYS: 3
|
408 |
+
OPTIMIZER: ADAMW
|
409 |
+
POLY_LR_CONSTANT_ENDING: 0.0
|
410 |
+
POLY_LR_POWER: 0.9
|
411 |
+
REFERENCE_WORLD_SIZE: 0
|
412 |
+
RESCALE_INTERVAL: false
|
413 |
+
STEPS:
|
414 |
+
- 8500
|
415 |
+
- 9500
|
416 |
+
WARMUP_FACTOR: 1.0
|
417 |
+
WARMUP_ITERS: 0
|
418 |
+
WARMUP_METHOD: linear
|
419 |
+
WEIGHT_DECAY: 0.05
|
420 |
+
WEIGHT_DECAY_BIAS: null
|
421 |
+
WEIGHT_DECAY_EMBED: 0.0
|
422 |
+
WEIGHT_DECAY_NORM: 0.0
|
423 |
+
TEST:
|
424 |
+
AUG:
|
425 |
+
ENABLED: false
|
426 |
+
FLIP: true
|
427 |
+
MAX_SIZE: 4480
|
428 |
+
MIN_SIZES:
|
429 |
+
- 320
|
430 |
+
- 480
|
431 |
+
- 640
|
432 |
+
- 800
|
433 |
+
- 960
|
434 |
+
- 1120
|
435 |
+
DETECTIONS_PER_IMAGE: 100
|
436 |
+
EVAL_PERIOD: 5000
|
437 |
+
EXPECTED_RESULTS: []
|
438 |
+
KEYPOINT_OKS_SIGMAS: []
|
439 |
+
PRECISE_BN:
|
440 |
+
ENABLED: false
|
441 |
+
NUM_ITER: 200
|
442 |
+
VERSION: 2
|
443 |
+
VIS_PERIOD: 0
|
ps_100-10_42/step6/fake_query.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e99a4240052c6dee1d381c6f6860c9dc597611cd061c0d7952952e7317bea65d
|
3 |
+
size 2485144137
|
ps_100-10_42/step6/last_checkpoint
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
model_final.pth
|
ps_100-10_42/step6/metrics.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
ps_100-10_42/step6/model_final.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7075a5e1d30c2dd88501569c09c49e61b99a804d78e3ea8ea07d1edbddf2c7b6
|
3 |
+
size 534485834
|
ps_100-10_42/step6/psd_distribution.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
[2047.0, 443.0, 459.0, 1522.0, 314.0, 1027.0, 215.0, 421.0, 1281.0, 102.0, 1268.0, 235.0, 2588.0, 19.0, 481.0, 1403.0, 35.0, 420.0, 724.0, 2526.0, 721.0, 32.0, 1348.0, 235.0, 226.0, 9.0, 24.0, 178.0, 172.0, 0.0, 285.0, 131.0, 26.0, 199.0, 3.0, 34.0, 767.0, 59.0, 37.0, 721.0, 0.0, 138.0, 166.0, 272.0, 62.0, 46.0, 2.0, 246.0, 22.0, 103.0, 52.0, 0.0, 4.0, 17.0, 1.0, 13.0, 35.0, 372.0, 26.0, 13.0, 1.0, 0.0, 57.0, 26.0, 141.0, 77.0, 308.0, 584.0, 0.0, 24.0, 58.0, 111.0, 37.0, 37.0, 118.0, 179.0, 61.0, 16.0, 15.0, 0.0, 11.0, 129.0, 1003.0, 9.0, 10.0, 140.0, 65.0, 127.0, 0.0, 65.0, 9.0, 0.0, 12.0, 54.0, 0.0, 5.0, 0.0, 22.0, 506.0, 9.0, 115.0, 8.0, 76.0, 9.0, 6.0, 12.0, 5.0, 10.0, 141.0, 14.0, 176.0, 2.0, 135.0, 0.0, 4.0, 56.0, 23.0, 7.0, 49.0, 17.0, 115.0, 59.0, 2.0, 69.0, 99.0, 320.0, 6.0, 26.0, 26.0, 64.0, 18.0, 21.0, 39.0, 63.0, 253.0, 309.0, 37.0, 67.0, 130.0, 58.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
|
ps_100-50_40/step2/config.yaml
ADDED
@@ -0,0 +1,439 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
CONT:
|
2 |
+
BASE_CLS: 100
|
3 |
+
COLLECT_QUERY_MODE: false
|
4 |
+
CUMULATIVE_PSDNUM: false
|
5 |
+
INC_CLS: 50
|
6 |
+
LIB_SIZE: 80
|
7 |
+
MED_TOKENS_WEIGHT: 5.0
|
8 |
+
MEMORY: false
|
9 |
+
OLD_WEIGHTS: ./output/ps/100-50_fake1_weakdist/step1/model_final.pth
|
10 |
+
PSD_LABEL_THRESHOLD: 0.35
|
11 |
+
SETTING: overlapped
|
12 |
+
TASK: 2
|
13 |
+
TOT_CLS: 150
|
14 |
+
VQ_NUMBER: 1
|
15 |
+
WEIGHTED_SAMPLE: true
|
16 |
+
WEIGHTS: null
|
17 |
+
CUDNN_BENCHMARK: false
|
18 |
+
DATALOADER:
|
19 |
+
ASPECT_RATIO_GROUPING: true
|
20 |
+
FILTER_EMPTY_ANNOTATIONS: true
|
21 |
+
NUM_WORKERS: 2
|
22 |
+
REPEAT_SQRT: true
|
23 |
+
REPEAT_THRESHOLD: 0.0
|
24 |
+
SAMPLER_TRAIN: TrainingSampler
|
25 |
+
DATASETS:
|
26 |
+
PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
|
27 |
+
PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
|
28 |
+
PROPOSAL_FILES_TEST: []
|
29 |
+
PROPOSAL_FILES_TRAIN: []
|
30 |
+
TEST:
|
31 |
+
- current_ade20k_panoptic_val
|
32 |
+
TRAIN:
|
33 |
+
- current_ade20k_panoptic_train
|
34 |
+
GLOBAL:
|
35 |
+
HACK: 1.0
|
36 |
+
INPUT:
|
37 |
+
COLOR_AUG_SSD: true
|
38 |
+
CROP:
|
39 |
+
ENABLED: true
|
40 |
+
SINGLE_CATEGORY_MAX_AREA: 1.0
|
41 |
+
SIZE:
|
42 |
+
- 640
|
43 |
+
- 640
|
44 |
+
TYPE: absolute
|
45 |
+
DATASET_MAPPER_NAME: continual_panoptic
|
46 |
+
FORMAT: RGB
|
47 |
+
IMAGE_SIZE: 1024
|
48 |
+
MASK_FORMAT: polygon
|
49 |
+
MAX_SCALE: 2.0
|
50 |
+
MAX_SIZE_TEST: 2560
|
51 |
+
MAX_SIZE_TRAIN: 2560
|
52 |
+
MIN_SCALE: 0.1
|
53 |
+
MIN_SIZE_TEST: 640
|
54 |
+
MIN_SIZE_TRAIN:
|
55 |
+
- 320
|
56 |
+
- 384
|
57 |
+
- 448
|
58 |
+
- 512
|
59 |
+
- 576
|
60 |
+
- 640
|
61 |
+
- 704
|
62 |
+
- 768
|
63 |
+
- 832
|
64 |
+
- 896
|
65 |
+
- 960
|
66 |
+
- 1024
|
67 |
+
- 1088
|
68 |
+
- 1152
|
69 |
+
- 1216
|
70 |
+
- 1280
|
71 |
+
MIN_SIZE_TRAIN_SAMPLING: choice
|
72 |
+
RANDOM_FLIP: horizontal
|
73 |
+
SIZE_DIVISIBILITY: 640
|
74 |
+
MODEL:
|
75 |
+
ANCHOR_GENERATOR:
|
76 |
+
ANGLES:
|
77 |
+
- - -90
|
78 |
+
- 0
|
79 |
+
- 90
|
80 |
+
ASPECT_RATIOS:
|
81 |
+
- - 0.5
|
82 |
+
- 1.0
|
83 |
+
- 2.0
|
84 |
+
NAME: DefaultAnchorGenerator
|
85 |
+
OFFSET: 0.0
|
86 |
+
SIZES:
|
87 |
+
- - 32
|
88 |
+
- 64
|
89 |
+
- 128
|
90 |
+
- 256
|
91 |
+
- 512
|
92 |
+
BACKBONE:
|
93 |
+
FREEZE_AT: 0
|
94 |
+
NAME: build_resnet_backbone
|
95 |
+
DEVICE: cuda
|
96 |
+
FPN:
|
97 |
+
FUSE_TYPE: sum
|
98 |
+
IN_FEATURES: []
|
99 |
+
NORM: ''
|
100 |
+
OUT_CHANNELS: 256
|
101 |
+
KEYPOINT_ON: false
|
102 |
+
LOAD_PROPOSALS: false
|
103 |
+
MASK_FORMER:
|
104 |
+
CLASS_WEIGHT: 4.0
|
105 |
+
CLIP_DIM: 768
|
106 |
+
DEC_LAYERS: 10
|
107 |
+
DEEP_SUPERVISION: true
|
108 |
+
DICE_WEIGHT: 5.0
|
109 |
+
DIM_FEEDFORWARD: 2048
|
110 |
+
DROPOUT: 0.0
|
111 |
+
ENC_LAYERS: 0
|
112 |
+
ENFORCE_INPUT_PROJ: false
|
113 |
+
HIDDEN_DIM: 256
|
114 |
+
IMPORTANCE_SAMPLE_RATIO: 0.75
|
115 |
+
MASK_WEIGHT: 5.0
|
116 |
+
NHEADS: 8
|
117 |
+
NO_OBJECT_WEIGHT: 0.1
|
118 |
+
NUM_OBJECT_QUERIES: 100
|
119 |
+
OVERSAMPLE_RATIO: 3.0
|
120 |
+
PRE_NORM: false
|
121 |
+
SIZE_DIVISIBILITY: 32
|
122 |
+
TEST:
|
123 |
+
INSTANCE_ON: true
|
124 |
+
OBJECT_MASK_THRESHOLD: 0.2
|
125 |
+
OVERLAP_THRESHOLD: 0.8
|
126 |
+
PANOPTIC_ON: true
|
127 |
+
SEMANTIC_ON: true
|
128 |
+
SEM_SEG_POSTPROCESSING_BEFORE_INFERENCE: false
|
129 |
+
TEXT_PATH: label_embedding/ade20k_150_laion768_all.npy
|
130 |
+
TRAIN_NUM_POINTS: 12544
|
131 |
+
TRANSFORMER_DECODER_NAME: MultiScaleMaskedTransformerDecoder
|
132 |
+
TRANSFORMER_IN_FEATURE: multi_scale_pixel_decoder
|
133 |
+
USE_TEXT_EMBEDDING: false
|
134 |
+
MASK_ON: false
|
135 |
+
META_ARCHITECTURE: MaskFormer
|
136 |
+
PANOPTIC_FPN:
|
137 |
+
COMBINE:
|
138 |
+
ENABLED: true
|
139 |
+
INSTANCES_CONFIDENCE_THRESH: 0.5
|
140 |
+
OVERLAP_THRESH: 0.5
|
141 |
+
STUFF_AREA_LIMIT: 4096
|
142 |
+
INSTANCE_LOSS_WEIGHT: 1.0
|
143 |
+
PIXEL_MEAN:
|
144 |
+
- 123.675
|
145 |
+
- 116.28
|
146 |
+
- 103.53
|
147 |
+
PIXEL_STD:
|
148 |
+
- 58.395
|
149 |
+
- 57.12
|
150 |
+
- 57.375
|
151 |
+
PROPOSAL_GENERATOR:
|
152 |
+
MIN_SIZE: 0
|
153 |
+
NAME: RPN
|
154 |
+
RESNETS:
|
155 |
+
DEFORM_MODULATED: false
|
156 |
+
DEFORM_NUM_GROUPS: 1
|
157 |
+
DEFORM_ON_PER_STAGE:
|
158 |
+
- false
|
159 |
+
- false
|
160 |
+
- false
|
161 |
+
- false
|
162 |
+
DEPTH: 50
|
163 |
+
NORM: FrozenBN
|
164 |
+
NUM_GROUPS: 1
|
165 |
+
OUT_FEATURES:
|
166 |
+
- res2
|
167 |
+
- res3
|
168 |
+
- res4
|
169 |
+
- res5
|
170 |
+
RES2_OUT_CHANNELS: 256
|
171 |
+
RES4_DILATION: 1
|
172 |
+
RES5_DILATION: 1
|
173 |
+
RES5_MULTI_GRID:
|
174 |
+
- 1
|
175 |
+
- 1
|
176 |
+
- 1
|
177 |
+
STEM_OUT_CHANNELS: 64
|
178 |
+
STEM_TYPE: basic
|
179 |
+
STRIDE_IN_1X1: false
|
180 |
+
WIDTH_PER_GROUP: 64
|
181 |
+
RETINANET:
|
182 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
183 |
+
BBOX_REG_WEIGHTS: &id002
|
184 |
+
- 1.0
|
185 |
+
- 1.0
|
186 |
+
- 1.0
|
187 |
+
- 1.0
|
188 |
+
FOCAL_LOSS_ALPHA: 0.25
|
189 |
+
FOCAL_LOSS_GAMMA: 2.0
|
190 |
+
IN_FEATURES:
|
191 |
+
- p3
|
192 |
+
- p4
|
193 |
+
- p5
|
194 |
+
- p6
|
195 |
+
- p7
|
196 |
+
IOU_LABELS:
|
197 |
+
- 0
|
198 |
+
- -1
|
199 |
+
- 1
|
200 |
+
IOU_THRESHOLDS:
|
201 |
+
- 0.4
|
202 |
+
- 0.5
|
203 |
+
NMS_THRESH_TEST: 0.5
|
204 |
+
NORM: ''
|
205 |
+
NUM_CLASSES: 80
|
206 |
+
NUM_CONVS: 4
|
207 |
+
PRIOR_PROB: 0.01
|
208 |
+
SCORE_THRESH_TEST: 0.05
|
209 |
+
SMOOTH_L1_LOSS_BETA: 0.1
|
210 |
+
TOPK_CANDIDATES_TEST: 1000
|
211 |
+
ROI_BOX_CASCADE_HEAD:
|
212 |
+
BBOX_REG_WEIGHTS:
|
213 |
+
- &id001
|
214 |
+
- 10.0
|
215 |
+
- 10.0
|
216 |
+
- 5.0
|
217 |
+
- 5.0
|
218 |
+
- - 20.0
|
219 |
+
- 20.0
|
220 |
+
- 10.0
|
221 |
+
- 10.0
|
222 |
+
- - 30.0
|
223 |
+
- 30.0
|
224 |
+
- 15.0
|
225 |
+
- 15.0
|
226 |
+
IOUS:
|
227 |
+
- 0.5
|
228 |
+
- 0.6
|
229 |
+
- 0.7
|
230 |
+
ROI_BOX_HEAD:
|
231 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
232 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
233 |
+
BBOX_REG_WEIGHTS: *id001
|
234 |
+
CLS_AGNOSTIC_BBOX_REG: false
|
235 |
+
CONV_DIM: 256
|
236 |
+
FC_DIM: 1024
|
237 |
+
FED_LOSS_FREQ_WEIGHT_POWER: 0.5
|
238 |
+
FED_LOSS_NUM_CLASSES: 50
|
239 |
+
NAME: ''
|
240 |
+
NORM: ''
|
241 |
+
NUM_CONV: 0
|
242 |
+
NUM_FC: 0
|
243 |
+
POOLER_RESOLUTION: 14
|
244 |
+
POOLER_SAMPLING_RATIO: 0
|
245 |
+
POOLER_TYPE: ROIAlignV2
|
246 |
+
SMOOTH_L1_BETA: 0.0
|
247 |
+
TRAIN_ON_PRED_BOXES: false
|
248 |
+
USE_FED_LOSS: false
|
249 |
+
USE_SIGMOID_CE: false
|
250 |
+
ROI_HEADS:
|
251 |
+
BATCH_SIZE_PER_IMAGE: 512
|
252 |
+
IN_FEATURES:
|
253 |
+
- res4
|
254 |
+
IOU_LABELS:
|
255 |
+
- 0
|
256 |
+
- 1
|
257 |
+
IOU_THRESHOLDS:
|
258 |
+
- 0.5
|
259 |
+
NAME: Res5ROIHeads
|
260 |
+
NMS_THRESH_TEST: 0.5
|
261 |
+
NUM_CLASSES: 80
|
262 |
+
POSITIVE_FRACTION: 0.25
|
263 |
+
PROPOSAL_APPEND_GT: true
|
264 |
+
SCORE_THRESH_TEST: 0.05
|
265 |
+
ROI_KEYPOINT_HEAD:
|
266 |
+
CONV_DIMS:
|
267 |
+
- 512
|
268 |
+
- 512
|
269 |
+
- 512
|
270 |
+
- 512
|
271 |
+
- 512
|
272 |
+
- 512
|
273 |
+
- 512
|
274 |
+
- 512
|
275 |
+
LOSS_WEIGHT: 1.0
|
276 |
+
MIN_KEYPOINTS_PER_IMAGE: 1
|
277 |
+
NAME: KRCNNConvDeconvUpsampleHead
|
278 |
+
NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
|
279 |
+
NUM_KEYPOINTS: 17
|
280 |
+
POOLER_RESOLUTION: 14
|
281 |
+
POOLER_SAMPLING_RATIO: 0
|
282 |
+
POOLER_TYPE: ROIAlignV2
|
283 |
+
ROI_MASK_HEAD:
|
284 |
+
CLS_AGNOSTIC_MASK: false
|
285 |
+
CONV_DIM: 256
|
286 |
+
NAME: MaskRCNNConvUpsampleHead
|
287 |
+
NORM: ''
|
288 |
+
NUM_CONV: 0
|
289 |
+
POOLER_RESOLUTION: 14
|
290 |
+
POOLER_SAMPLING_RATIO: 0
|
291 |
+
POOLER_TYPE: ROIAlignV2
|
292 |
+
RPN:
|
293 |
+
BATCH_SIZE_PER_IMAGE: 256
|
294 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
295 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
296 |
+
BBOX_REG_WEIGHTS: *id002
|
297 |
+
BOUNDARY_THRESH: -1
|
298 |
+
CONV_DIMS:
|
299 |
+
- -1
|
300 |
+
HEAD_NAME: StandardRPNHead
|
301 |
+
IN_FEATURES:
|
302 |
+
- res4
|
303 |
+
IOU_LABELS:
|
304 |
+
- 0
|
305 |
+
- -1
|
306 |
+
- 1
|
307 |
+
IOU_THRESHOLDS:
|
308 |
+
- 0.3
|
309 |
+
- 0.7
|
310 |
+
LOSS_WEIGHT: 1.0
|
311 |
+
NMS_THRESH: 0.7
|
312 |
+
POSITIVE_FRACTION: 0.5
|
313 |
+
POST_NMS_TOPK_TEST: 1000
|
314 |
+
POST_NMS_TOPK_TRAIN: 2000
|
315 |
+
PRE_NMS_TOPK_TEST: 6000
|
316 |
+
PRE_NMS_TOPK_TRAIN: 12000
|
317 |
+
SMOOTH_L1_BETA: 0.0
|
318 |
+
SEM_SEG_HEAD:
|
319 |
+
ASPP_CHANNELS: 256
|
320 |
+
ASPP_DILATIONS:
|
321 |
+
- 6
|
322 |
+
- 12
|
323 |
+
- 18
|
324 |
+
ASPP_DROPOUT: 0.1
|
325 |
+
COMMON_STRIDE: 4
|
326 |
+
CONVS_DIM: 256
|
327 |
+
DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES:
|
328 |
+
- res3
|
329 |
+
- res4
|
330 |
+
- res5
|
331 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_HEADS: 8
|
332 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_POINTS: 4
|
333 |
+
IGNORE_VALUE: 255
|
334 |
+
IN_FEATURES:
|
335 |
+
- res2
|
336 |
+
- res3
|
337 |
+
- res4
|
338 |
+
- res5
|
339 |
+
LOSS_TYPE: hard_pixel_mining
|
340 |
+
LOSS_WEIGHT: 1.0
|
341 |
+
MASK_DIM: 256
|
342 |
+
NAME: MaskFormerHead
|
343 |
+
NORM: GN
|
344 |
+
NUM_CLASSES: 150
|
345 |
+
PIXEL_DECODER_NAME: MSDeformAttnPixelDecoder
|
346 |
+
PROJECT_CHANNELS:
|
347 |
+
- 48
|
348 |
+
PROJECT_FEATURES:
|
349 |
+
- res2
|
350 |
+
TRANSFORMER_ENC_LAYERS: 6
|
351 |
+
USE_DEPTHWISE_SEPARABLE_CONV: false
|
352 |
+
SWIN:
|
353 |
+
APE: false
|
354 |
+
ATTN_DROP_RATE: 0.0
|
355 |
+
DEPTHS:
|
356 |
+
- 2
|
357 |
+
- 2
|
358 |
+
- 6
|
359 |
+
- 2
|
360 |
+
DROP_PATH_RATE: 0.3
|
361 |
+
DROP_RATE: 0.0
|
362 |
+
EMBED_DIM: 96
|
363 |
+
MLP_RATIO: 4.0
|
364 |
+
NUM_HEADS:
|
365 |
+
- 3
|
366 |
+
- 6
|
367 |
+
- 12
|
368 |
+
- 24
|
369 |
+
OUT_FEATURES:
|
370 |
+
- res2
|
371 |
+
- res3
|
372 |
+
- res4
|
373 |
+
- res5
|
374 |
+
PATCH_NORM: true
|
375 |
+
PATCH_SIZE: 4
|
376 |
+
PRETRAIN_IMG_SIZE: 224
|
377 |
+
QKV_BIAS: true
|
378 |
+
QK_SCALE: null
|
379 |
+
USE_CHECKPOINT: false
|
380 |
+
WINDOW_SIZE: 7
|
381 |
+
WEIGHTS: ./output/ps/100-50_fake1_weakdist/step1/model_final.pth
|
382 |
+
OUTPUT_DIR: ./output/ps/100-50_fake1_weakdist/step2
|
383 |
+
SEED: -1
|
384 |
+
SOLVER:
|
385 |
+
AMP:
|
386 |
+
ENABLED: true
|
387 |
+
BACKBONE_MULTIPLIER: 0.1
|
388 |
+
BASE_LR: 5.0e-05
|
389 |
+
BASE_LR_END: 0.0
|
390 |
+
BIAS_LR_FACTOR: 1.0
|
391 |
+
CHECKPOINT_PERIOD: 10000
|
392 |
+
CLIP_GRADIENTS:
|
393 |
+
CLIP_TYPE: full_model
|
394 |
+
CLIP_VALUE: 0.01
|
395 |
+
ENABLED: true
|
396 |
+
NORM_TYPE: 2.0
|
397 |
+
GAMMA: 0.1
|
398 |
+
IMS_PER_BATCH: 8
|
399 |
+
LR_SCHEDULER_NAME: WarmupMultiStepLR
|
400 |
+
MAX_ITER: 50000
|
401 |
+
MOMENTUM: 0.9
|
402 |
+
NESTEROV: false
|
403 |
+
NUM_DECAYS: 3
|
404 |
+
OPTIMIZER: ADAMW
|
405 |
+
POLY_LR_CONSTANT_ENDING: 0.0
|
406 |
+
POLY_LR_POWER: 0.9
|
407 |
+
REFERENCE_WORLD_SIZE: 0
|
408 |
+
RESCALE_INTERVAL: false
|
409 |
+
STEPS:
|
410 |
+
- 42500
|
411 |
+
- 47500
|
412 |
+
WARMUP_FACTOR: 1.0
|
413 |
+
WARMUP_ITERS: 0
|
414 |
+
WARMUP_METHOD: linear
|
415 |
+
WEIGHT_DECAY: 0.05
|
416 |
+
WEIGHT_DECAY_BIAS: null
|
417 |
+
WEIGHT_DECAY_EMBED: 0.0
|
418 |
+
WEIGHT_DECAY_NORM: 0.0
|
419 |
+
TEST:
|
420 |
+
AUG:
|
421 |
+
ENABLED: false
|
422 |
+
FLIP: true
|
423 |
+
MAX_SIZE: 4480
|
424 |
+
MIN_SIZES:
|
425 |
+
- 320
|
426 |
+
- 480
|
427 |
+
- 640
|
428 |
+
- 800
|
429 |
+
- 960
|
430 |
+
- 1120
|
431 |
+
DETECTIONS_PER_IMAGE: 100
|
432 |
+
EVAL_PERIOD: 5000
|
433 |
+
EXPECTED_RESULTS: []
|
434 |
+
KEYPOINT_OKS_SIGMAS: []
|
435 |
+
PRECISE_BN:
|
436 |
+
ENABLED: false
|
437 |
+
NUM_ITER: 200
|
438 |
+
VERSION: 2
|
439 |
+
VIS_PERIOD: 0
|
ps_100-50_40/step2/last_checkpoint
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
model_final.pth
|
ps_100-50_40/step2/metrics.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
ps_100-50_40/step2/model_final.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:79776abf7639435b10ff03e67dc324e443874ae9a461a57bf973e05750cbd879
|
3 |
+
size 534485834
|
ps_100-50_40/step2/psd_distribution.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
[4002.0, 8077.0, 7490.0, 2601.0, 4987.0, 2168.0, 6027.0, 410.0, 1326.0, 877.0, 343.0, 4170.0, 19436.0, 143.0, 836.0, 1948.0, 849.0, 1158.0, 981.0, 4268.0, 24375.0, 136.0, 1563.0, 173.0, 487.0, 45.0, 315.0, 135.0, 150.0, 50.0, 180.0, 1731.0, 509.0, 443.0, 43.0, 14.0, 587.0, 0.0, 124.0, 406.0, 43.0, 299.0, 426.0, 3686.0, 35.0, 68.0, 10.0, 54.0, 159.0, 20.0, 21.0, 33.0, 31.0, 271.0, 109.0, 68.0, 65.0, 203.0, 0.0, 74.0, 9.0, 103.0, 181.0, 19.0, 60.0, 12.0, 178.0, 1276.0, 6.0, 337.0, 0.0, 45.0, 639.0, 1.0, 440.0, 624.0, 174.0, 21.0, 33.0, 0.0, 423.0, 0.0, 2975.0, 540.0, 84.0, 117.0, 877.0, 2908.0, 16.0, 88.0, 337.0, 1.0, 29.0, 561.0, 16.0, 76.0, 11.0, 12.0, 199.0, 12.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
|
ps_100-5_35/step11/config.yaml
ADDED
@@ -0,0 +1,452 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
CONT:
|
2 |
+
ADD_POS: false
|
3 |
+
BASE_CLS: 100
|
4 |
+
COLLECT_QUERY_MODE: false
|
5 |
+
COMBINE_PSDLABEL: false
|
6 |
+
CUMULATIVE_PSDNUM: false
|
7 |
+
DISTRIBUTION_ALPHA: 0.5
|
8 |
+
FILTER_KD: false
|
9 |
+
FREEZE_LABEL: false
|
10 |
+
INC_CLS: 5
|
11 |
+
KD_DECODER: false
|
12 |
+
KD_TEMPERATURE: 1.0
|
13 |
+
KD_TEMPERATURE2: 1.0
|
14 |
+
KD_TYPE: kl
|
15 |
+
KL_ALL: true
|
16 |
+
KL_WEIGHT: 2.0
|
17 |
+
LIB_SIZE: 80
|
18 |
+
MED_TOKENS_WEIGHT: 5.0
|
19 |
+
MEMORY: false
|
20 |
+
OLD_WEIGHTS: /public/home/zhuyuchen530/projects/cvpr24/fuke/fake3_clone/output/ps/pan_100-5_FL/step11/model_final.pth
|
21 |
+
PSD_LABEL_THRESHOLD: 0.35
|
22 |
+
PSD_OVERLAP_THRESHOLD: 0.8
|
23 |
+
SETTING: overlapped
|
24 |
+
TASK: 11
|
25 |
+
TOT_CLS: 150
|
26 |
+
VQ_NUMBER: 5
|
27 |
+
VQ_STORE: false
|
28 |
+
WEIGHTED_SAMPLE: true
|
29 |
+
WEIGHTS: null
|
30 |
+
CUDNN_BENCHMARK: false
|
31 |
+
DATALOADER:
|
32 |
+
ASPECT_RATIO_GROUPING: true
|
33 |
+
FILTER_EMPTY_ANNOTATIONS: true
|
34 |
+
NUM_WORKERS: 2
|
35 |
+
REPEAT_SQRT: true
|
36 |
+
REPEAT_THRESHOLD: 0.0
|
37 |
+
SAMPLER_TRAIN: TrainingSampler
|
38 |
+
DATASETS:
|
39 |
+
PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
|
40 |
+
PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
|
41 |
+
PROPOSAL_FILES_TEST: []
|
42 |
+
PROPOSAL_FILES_TRAIN: []
|
43 |
+
TEST:
|
44 |
+
- current_ade20k_panoptic_val
|
45 |
+
TRAIN:
|
46 |
+
- current_ade20k_panoptic_train
|
47 |
+
GLOBAL:
|
48 |
+
HACK: 1.0
|
49 |
+
INPUT:
|
50 |
+
COLOR_AUG_SSD: true
|
51 |
+
CROP:
|
52 |
+
ENABLED: true
|
53 |
+
SINGLE_CATEGORY_MAX_AREA: 1.0
|
54 |
+
SIZE:
|
55 |
+
- 640
|
56 |
+
- 640
|
57 |
+
TYPE: absolute
|
58 |
+
DATASET_MAPPER_NAME: continual_panoptic
|
59 |
+
FORMAT: RGB
|
60 |
+
IMAGE_SIZE: 1024
|
61 |
+
MASK_FORMAT: polygon
|
62 |
+
MAX_SCALE: 2.0
|
63 |
+
MAX_SIZE_TEST: 2560
|
64 |
+
MAX_SIZE_TRAIN: 2560
|
65 |
+
MIN_SCALE: 0.1
|
66 |
+
MIN_SIZE_TEST: 640
|
67 |
+
MIN_SIZE_TRAIN:
|
68 |
+
- 320
|
69 |
+
- 384
|
70 |
+
- 448
|
71 |
+
- 512
|
72 |
+
- 576
|
73 |
+
- 640
|
74 |
+
- 704
|
75 |
+
- 768
|
76 |
+
- 832
|
77 |
+
- 896
|
78 |
+
- 960
|
79 |
+
- 1024
|
80 |
+
- 1088
|
81 |
+
- 1152
|
82 |
+
- 1216
|
83 |
+
- 1280
|
84 |
+
MIN_SIZE_TRAIN_SAMPLING: choice
|
85 |
+
RANDOM_FLIP: horizontal
|
86 |
+
SIZE_DIVISIBILITY: 640
|
87 |
+
MODEL:
|
88 |
+
ANCHOR_GENERATOR:
|
89 |
+
ANGLES:
|
90 |
+
- - -90
|
91 |
+
- 0
|
92 |
+
- 90
|
93 |
+
ASPECT_RATIOS:
|
94 |
+
- - 0.5
|
95 |
+
- 1.0
|
96 |
+
- 2.0
|
97 |
+
NAME: DefaultAnchorGenerator
|
98 |
+
OFFSET: 0.0
|
99 |
+
SIZES:
|
100 |
+
- - 32
|
101 |
+
- 64
|
102 |
+
- 128
|
103 |
+
- 256
|
104 |
+
- 512
|
105 |
+
BACKBONE:
|
106 |
+
FREEZE_AT: 0
|
107 |
+
NAME: build_resnet_backbone
|
108 |
+
DEVICE: cuda
|
109 |
+
FPN:
|
110 |
+
FUSE_TYPE: sum
|
111 |
+
IN_FEATURES: []
|
112 |
+
NORM: ''
|
113 |
+
OUT_CHANNELS: 256
|
114 |
+
KEYPOINT_ON: false
|
115 |
+
LOAD_PROPOSALS: false
|
116 |
+
MASK_FORMER:
|
117 |
+
CLASS_WEIGHT: 4.0
|
118 |
+
CLIP_DIM: 768
|
119 |
+
DEC_LAYERS: 10
|
120 |
+
DEEP_SUPERVISION: true
|
121 |
+
DICE_WEIGHT: 5.0
|
122 |
+
DIM_FEEDFORWARD: 2048
|
123 |
+
DROPOUT: 0.0
|
124 |
+
ENC_LAYERS: 0
|
125 |
+
ENFORCE_INPUT_PROJ: false
|
126 |
+
HIDDEN_DIM: 256
|
127 |
+
IMPORTANCE_SAMPLE_RATIO: 0.75
|
128 |
+
MASK_WEIGHT: 5.0
|
129 |
+
NHEADS: 8
|
130 |
+
NO_OBJECT_WEIGHT: 0.1
|
131 |
+
NUM_OBJECT_QUERIES: 100
|
132 |
+
OVERSAMPLE_RATIO: 3.0
|
133 |
+
PRE_NORM: false
|
134 |
+
SIZE_DIVISIBILITY: 32
|
135 |
+
TEST:
|
136 |
+
INSTANCE_ON: true
|
137 |
+
OBJECT_MASK_THRESHOLD: 0.33
|
138 |
+
OVERLAP_THRESHOLD: 0.8
|
139 |
+
PANOPTIC_ON: true
|
140 |
+
SEMANTIC_ON: true
|
141 |
+
SEM_SEG_POSTPROCESSING_BEFORE_INFERENCE: false
|
142 |
+
TEXT_PATH: label_embedding/ade20k_150_laion768_all.npy
|
143 |
+
TRAIN_NUM_POINTS: 12544
|
144 |
+
TRANSFORMER_DECODER_NAME: MultiScaleMaskedTransformerDecoder
|
145 |
+
TRANSFORMER_IN_FEATURE: multi_scale_pixel_decoder
|
146 |
+
USE_TEXT_EMBEDDING: false
|
147 |
+
MASK_ON: false
|
148 |
+
META_ARCHITECTURE: MaskFormer
|
149 |
+
PANOPTIC_FPN:
|
150 |
+
COMBINE:
|
151 |
+
ENABLED: true
|
152 |
+
INSTANCES_CONFIDENCE_THRESH: 0.5
|
153 |
+
OVERLAP_THRESH: 0.5
|
154 |
+
STUFF_AREA_LIMIT: 4096
|
155 |
+
INSTANCE_LOSS_WEIGHT: 1.0
|
156 |
+
PIXEL_MEAN:
|
157 |
+
- 123.675
|
158 |
+
- 116.28
|
159 |
+
- 103.53
|
160 |
+
PIXEL_STD:
|
161 |
+
- 58.395
|
162 |
+
- 57.12
|
163 |
+
- 57.375
|
164 |
+
PROPOSAL_GENERATOR:
|
165 |
+
MIN_SIZE: 0
|
166 |
+
NAME: RPN
|
167 |
+
RESNETS:
|
168 |
+
DEFORM_MODULATED: false
|
169 |
+
DEFORM_NUM_GROUPS: 1
|
170 |
+
DEFORM_ON_PER_STAGE:
|
171 |
+
- false
|
172 |
+
- false
|
173 |
+
- false
|
174 |
+
- false
|
175 |
+
DEPTH: 50
|
176 |
+
NORM: FrozenBN
|
177 |
+
NUM_GROUPS: 1
|
178 |
+
OUT_FEATURES:
|
179 |
+
- res2
|
180 |
+
- res3
|
181 |
+
- res4
|
182 |
+
- res5
|
183 |
+
RES2_OUT_CHANNELS: 256
|
184 |
+
RES4_DILATION: 1
|
185 |
+
RES5_DILATION: 1
|
186 |
+
RES5_MULTI_GRID:
|
187 |
+
- 1
|
188 |
+
- 1
|
189 |
+
- 1
|
190 |
+
STEM_OUT_CHANNELS: 64
|
191 |
+
STEM_TYPE: basic
|
192 |
+
STRIDE_IN_1X1: false
|
193 |
+
WIDTH_PER_GROUP: 64
|
194 |
+
RETINANET:
|
195 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
196 |
+
BBOX_REG_WEIGHTS: &id002
|
197 |
+
- 1.0
|
198 |
+
- 1.0
|
199 |
+
- 1.0
|
200 |
+
- 1.0
|
201 |
+
FOCAL_LOSS_ALPHA: 0.25
|
202 |
+
FOCAL_LOSS_GAMMA: 2.0
|
203 |
+
IN_FEATURES:
|
204 |
+
- p3
|
205 |
+
- p4
|
206 |
+
- p5
|
207 |
+
- p6
|
208 |
+
- p7
|
209 |
+
IOU_LABELS:
|
210 |
+
- 0
|
211 |
+
- -1
|
212 |
+
- 1
|
213 |
+
IOU_THRESHOLDS:
|
214 |
+
- 0.4
|
215 |
+
- 0.5
|
216 |
+
NMS_THRESH_TEST: 0.5
|
217 |
+
NORM: ''
|
218 |
+
NUM_CLASSES: 80
|
219 |
+
NUM_CONVS: 4
|
220 |
+
PRIOR_PROB: 0.01
|
221 |
+
SCORE_THRESH_TEST: 0.05
|
222 |
+
SMOOTH_L1_LOSS_BETA: 0.1
|
223 |
+
TOPK_CANDIDATES_TEST: 1000
|
224 |
+
ROI_BOX_CASCADE_HEAD:
|
225 |
+
BBOX_REG_WEIGHTS:
|
226 |
+
- &id001
|
227 |
+
- 10.0
|
228 |
+
- 10.0
|
229 |
+
- 5.0
|
230 |
+
- 5.0
|
231 |
+
- - 20.0
|
232 |
+
- 20.0
|
233 |
+
- 10.0
|
234 |
+
- 10.0
|
235 |
+
- - 30.0
|
236 |
+
- 30.0
|
237 |
+
- 15.0
|
238 |
+
- 15.0
|
239 |
+
IOUS:
|
240 |
+
- 0.5
|
241 |
+
- 0.6
|
242 |
+
- 0.7
|
243 |
+
ROI_BOX_HEAD:
|
244 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
245 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
246 |
+
BBOX_REG_WEIGHTS: *id001
|
247 |
+
CLS_AGNOSTIC_BBOX_REG: false
|
248 |
+
CONV_DIM: 256
|
249 |
+
FC_DIM: 1024
|
250 |
+
FED_LOSS_FREQ_WEIGHT_POWER: 0.5
|
251 |
+
FED_LOSS_NUM_CLASSES: 50
|
252 |
+
NAME: ''
|
253 |
+
NORM: ''
|
254 |
+
NUM_CONV: 0
|
255 |
+
NUM_FC: 0
|
256 |
+
POOLER_RESOLUTION: 14
|
257 |
+
POOLER_SAMPLING_RATIO: 0
|
258 |
+
POOLER_TYPE: ROIAlignV2
|
259 |
+
SMOOTH_L1_BETA: 0.0
|
260 |
+
TRAIN_ON_PRED_BOXES: false
|
261 |
+
USE_FED_LOSS: false
|
262 |
+
USE_SIGMOID_CE: false
|
263 |
+
ROI_HEADS:
|
264 |
+
BATCH_SIZE_PER_IMAGE: 512
|
265 |
+
IN_FEATURES:
|
266 |
+
- res4
|
267 |
+
IOU_LABELS:
|
268 |
+
- 0
|
269 |
+
- 1
|
270 |
+
IOU_THRESHOLDS:
|
271 |
+
- 0.5
|
272 |
+
NAME: Res5ROIHeads
|
273 |
+
NMS_THRESH_TEST: 0.5
|
274 |
+
NUM_CLASSES: 80
|
275 |
+
POSITIVE_FRACTION: 0.25
|
276 |
+
PROPOSAL_APPEND_GT: true
|
277 |
+
SCORE_THRESH_TEST: 0.05
|
278 |
+
ROI_KEYPOINT_HEAD:
|
279 |
+
CONV_DIMS:
|
280 |
+
- 512
|
281 |
+
- 512
|
282 |
+
- 512
|
283 |
+
- 512
|
284 |
+
- 512
|
285 |
+
- 512
|
286 |
+
- 512
|
287 |
+
- 512
|
288 |
+
LOSS_WEIGHT: 1.0
|
289 |
+
MIN_KEYPOINTS_PER_IMAGE: 1
|
290 |
+
NAME: KRCNNConvDeconvUpsampleHead
|
291 |
+
NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
|
292 |
+
NUM_KEYPOINTS: 17
|
293 |
+
POOLER_RESOLUTION: 14
|
294 |
+
POOLER_SAMPLING_RATIO: 0
|
295 |
+
POOLER_TYPE: ROIAlignV2
|
296 |
+
ROI_MASK_HEAD:
|
297 |
+
CLS_AGNOSTIC_MASK: false
|
298 |
+
CONV_DIM: 256
|
299 |
+
NAME: MaskRCNNConvUpsampleHead
|
300 |
+
NORM: ''
|
301 |
+
NUM_CONV: 0
|
302 |
+
POOLER_RESOLUTION: 14
|
303 |
+
POOLER_SAMPLING_RATIO: 0
|
304 |
+
POOLER_TYPE: ROIAlignV2
|
305 |
+
RPN:
|
306 |
+
BATCH_SIZE_PER_IMAGE: 256
|
307 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
308 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
309 |
+
BBOX_REG_WEIGHTS: *id002
|
310 |
+
BOUNDARY_THRESH: -1
|
311 |
+
CONV_DIMS:
|
312 |
+
- -1
|
313 |
+
HEAD_NAME: StandardRPNHead
|
314 |
+
IN_FEATURES:
|
315 |
+
- res4
|
316 |
+
IOU_LABELS:
|
317 |
+
- 0
|
318 |
+
- -1
|
319 |
+
- 1
|
320 |
+
IOU_THRESHOLDS:
|
321 |
+
- 0.3
|
322 |
+
- 0.7
|
323 |
+
LOSS_WEIGHT: 1.0
|
324 |
+
NMS_THRESH: 0.7
|
325 |
+
POSITIVE_FRACTION: 0.5
|
326 |
+
POST_NMS_TOPK_TEST: 1000
|
327 |
+
POST_NMS_TOPK_TRAIN: 2000
|
328 |
+
PRE_NMS_TOPK_TEST: 6000
|
329 |
+
PRE_NMS_TOPK_TRAIN: 12000
|
330 |
+
SMOOTH_L1_BETA: 0.0
|
331 |
+
SEM_SEG_HEAD:
|
332 |
+
ASPP_CHANNELS: 256
|
333 |
+
ASPP_DILATIONS:
|
334 |
+
- 6
|
335 |
+
- 12
|
336 |
+
- 18
|
337 |
+
ASPP_DROPOUT: 0.1
|
338 |
+
COMMON_STRIDE: 4
|
339 |
+
CONVS_DIM: 256
|
340 |
+
DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES:
|
341 |
+
- res3
|
342 |
+
- res4
|
343 |
+
- res5
|
344 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_HEADS: 8
|
345 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_POINTS: 4
|
346 |
+
IGNORE_VALUE: 255
|
347 |
+
IN_FEATURES:
|
348 |
+
- res2
|
349 |
+
- res3
|
350 |
+
- res4
|
351 |
+
- res5
|
352 |
+
LOSS_TYPE: hard_pixel_mining
|
353 |
+
LOSS_WEIGHT: 1.0
|
354 |
+
MASK_DIM: 256
|
355 |
+
NAME: MaskFormerHead
|
356 |
+
NORM: GN
|
357 |
+
NUM_CLASSES: 150
|
358 |
+
PIXEL_DECODER_NAME: MSDeformAttnPixelDecoder
|
359 |
+
PROJECT_CHANNELS:
|
360 |
+
- 48
|
361 |
+
PROJECT_FEATURES:
|
362 |
+
- res2
|
363 |
+
TRANSFORMER_ENC_LAYERS: 6
|
364 |
+
USE_DEPTHWISE_SEPARABLE_CONV: false
|
365 |
+
SWIN:
|
366 |
+
APE: false
|
367 |
+
ATTN_DROP_RATE: 0.0
|
368 |
+
DEPTHS:
|
369 |
+
- 2
|
370 |
+
- 2
|
371 |
+
- 6
|
372 |
+
- 2
|
373 |
+
DROP_PATH_RATE: 0.3
|
374 |
+
DROP_RATE: 0.0
|
375 |
+
EMBED_DIM: 96
|
376 |
+
MLP_RATIO: 4.0
|
377 |
+
NUM_HEADS:
|
378 |
+
- 3
|
379 |
+
- 6
|
380 |
+
- 12
|
381 |
+
- 24
|
382 |
+
OUT_FEATURES:
|
383 |
+
- res2
|
384 |
+
- res3
|
385 |
+
- res4
|
386 |
+
- res5
|
387 |
+
PATCH_NORM: true
|
388 |
+
PATCH_SIZE: 4
|
389 |
+
PRETRAIN_IMG_SIZE: 224
|
390 |
+
QKV_BIAS: true
|
391 |
+
QK_SCALE: null
|
392 |
+
USE_CHECKPOINT: false
|
393 |
+
WINDOW_SIZE: 7
|
394 |
+
WEIGHTS: /public/home/zhuyuchen530/projects/cvpr24/fuke/fake3_clone/output/ps/pan_100-5_FL/step11/model_final.pth
|
395 |
+
OUTPUT_DIR: /public/home/zhuyuchen530/projects/cvpr24/fuke/fake3_clone/output/ps/pan_100-5_FL/step11
|
396 |
+
SEED: -1
|
397 |
+
SOLVER:
|
398 |
+
AMP:
|
399 |
+
ENABLED: true
|
400 |
+
BACKBONE_MULTIPLIER: 0.1
|
401 |
+
BASE_LR: 5.0e-05
|
402 |
+
BASE_LR_END: 0.0
|
403 |
+
BIAS_LR_FACTOR: 1.0
|
404 |
+
CHECKPOINT_PERIOD: 10000
|
405 |
+
CLIP_GRADIENTS:
|
406 |
+
CLIP_TYPE: full_model
|
407 |
+
CLIP_VALUE: 0.01
|
408 |
+
ENABLED: true
|
409 |
+
NORM_TYPE: 2.0
|
410 |
+
GAMMA: 0.1
|
411 |
+
IMS_PER_BATCH: 8
|
412 |
+
LR_SCHEDULER_NAME: WarmupMultiStepLR
|
413 |
+
MAX_ITER: 5000
|
414 |
+
MOMENTUM: 0.9
|
415 |
+
NESTEROV: false
|
416 |
+
NUM_DECAYS: 3
|
417 |
+
OPTIMIZER: ADAMW
|
418 |
+
POLY_LR_CONSTANT_ENDING: 0.0
|
419 |
+
POLY_LR_POWER: 0.9
|
420 |
+
REFERENCE_WORLD_SIZE: 0
|
421 |
+
RESCALE_INTERVAL: false
|
422 |
+
STEPS:
|
423 |
+
- 4250
|
424 |
+
- 4750
|
425 |
+
WARMUP_FACTOR: 1.0
|
426 |
+
WARMUP_ITERS: 0
|
427 |
+
WARMUP_METHOD: linear
|
428 |
+
WEIGHT_DECAY: 0.05
|
429 |
+
WEIGHT_DECAY_BIAS: null
|
430 |
+
WEIGHT_DECAY_EMBED: 0.0
|
431 |
+
WEIGHT_DECAY_NORM: 0.0
|
432 |
+
TEST:
|
433 |
+
AUG:
|
434 |
+
ENABLED: false
|
435 |
+
FLIP: true
|
436 |
+
MAX_SIZE: 4480
|
437 |
+
MIN_SIZES:
|
438 |
+
- 320
|
439 |
+
- 480
|
440 |
+
- 640
|
441 |
+
- 800
|
442 |
+
- 960
|
443 |
+
- 1120
|
444 |
+
DETECTIONS_PER_IMAGE: 100
|
445 |
+
EVAL_PERIOD: 2501
|
446 |
+
EXPECTED_RESULTS: []
|
447 |
+
KEYPOINT_OKS_SIGMAS: []
|
448 |
+
PRECISE_BN:
|
449 |
+
ENABLED: false
|
450 |
+
NUM_ITER: 200
|
451 |
+
VERSION: 2
|
452 |
+
VIS_PERIOD: 0
|
ps_100-5_35/step11/fake_query.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:722cb87dcc7cc82061abc189fbb3ea2b50604c66a91bc649f42dc89e27e3e6e7
|
3 |
+
size 27746927
|
ps_100-5_35/step11/last_checkpoint
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
model_final.pth
|
ps_100-5_35/step11/metrics.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
ps_100-5_35/step11/model_final.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:550f7801b702659f619e996598e64948939e394d3ca1b06e889f70cbba82b77a
|
3 |
+
size 534485834
|
ps_100-5_35/step11/psd_distribution.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
[1330.0, 379.0, 359.0, 967.0, 263.0, 666.0, 197.0, 359.0, 841.0, 84.0, 725.0, 180.0, 1573.0, 5.0, 272.0, 952.0, 20.0, 256.0, 540.0, 1537.0, 649.0, 3.0, 1001.0, 177.0, 71.0, 3.0, 10.0, 144.0, 129.0, 0.0, 151.0, 44.0, 19.0, 48.0, 0.0, 30.0, 600.0, 65.0, 9.0, 535.0, 0.0, 58.0, 64.0, 145.0, 51.0, 2.0, 1.0, 165.0, 7.0, 73.0, 31.0, 0.0, 0.0, 8.0, 1.0, 0.0, 23.0, 304.0, 28.0, 5.0, 1.0, 0.0, 30.0, 26.0, 93.0, 69.0, 211.0, 262.0, 0.0, 4.0, 50.0, 61.0, 22.0, 24.0, 35.0, 79.0, 9.0, 4.0, 15.0, 0.0, 12.0, 118.0, 559.0, 7.0, 11.0, 74.0, 63.0, 115.0, 0.0, 44.0, 1.0, 0.0, 4.0, 42.0, 3.0, 2.0, 0.0, 16.0, 351.0, 13.0, 4.0, 0.0, 16.0, 0.0, 0.0, 0.0, 1.0, 1.0, 18.0, 14.0, 98.0, 0.0, 37.0, 0.0, 0.0, 3.0, 5.0, 4.0, 20.0, 1.0, 13.0, 39.0, 0.0, 50.0, 42.0, 180.0, 2.0, 11.0, 0.0, 33.0, 7.0, 1.0, 8.0, 22.0, 158.0, 201.0, 22.0, 0.0, 35.0, 22.0, 0.0, 1.0, 400.0, 45.0, 27.0, 0.0, 0.0, 0.0, 0.0, 0.0]
|
ss_100-10_42/step6/config.yaml
ADDED
@@ -0,0 +1,452 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
CONT:
|
2 |
+
ADD_POS: false
|
3 |
+
BASE_CLS: 100
|
4 |
+
COLLECT_QUERY_MODE: false
|
5 |
+
COMBINE_PSDLABEL: true
|
6 |
+
CUMULATIVE_PSDNUM: false
|
7 |
+
DISTRIBUTION_ALPHA: 0.5
|
8 |
+
FILTER_KD: false
|
9 |
+
FREEZE_LABEL: false
|
10 |
+
INC_CLS: 10
|
11 |
+
KD_DECODER: true
|
12 |
+
KD_TEMPERATURE: 1.0
|
13 |
+
KD_TEMPERATURE2: 4.0
|
14 |
+
KD_TYPE: kl
|
15 |
+
KL_ALL: true
|
16 |
+
KL_WEIGHT: 2.0
|
17 |
+
LIB_SIZE: 80
|
18 |
+
MED_TOKENS_WEIGHT: 5.0
|
19 |
+
MEMORY: false
|
20 |
+
OLD_WEIGHTS: ./output/newss/100-10_1_1+4_vq1/step5/model_final.pth
|
21 |
+
PSD_LABEL_THRESHOLD: 0.0
|
22 |
+
PSD_OVERLAP_THRESHOLD: 0.8
|
23 |
+
SETTING: overlapped
|
24 |
+
TASK: 6
|
25 |
+
TOT_CLS: 150
|
26 |
+
VQ_NUMBER: 1
|
27 |
+
VQ_STORE: false
|
28 |
+
WEIGHTED_SAMPLE: true
|
29 |
+
WEIGHTS: null
|
30 |
+
CUDNN_BENCHMARK: false
|
31 |
+
DATALOADER:
|
32 |
+
ASPECT_RATIO_GROUPING: true
|
33 |
+
FILTER_EMPTY_ANNOTATIONS: true
|
34 |
+
NUM_WORKERS: 2
|
35 |
+
REPEAT_SQRT: true
|
36 |
+
REPEAT_THRESHOLD: 0.0
|
37 |
+
SAMPLER_TRAIN: TrainingSampler
|
38 |
+
DATASETS:
|
39 |
+
PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
|
40 |
+
PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
|
41 |
+
PROPOSAL_FILES_TEST: []
|
42 |
+
PROPOSAL_FILES_TRAIN: []
|
43 |
+
TEST:
|
44 |
+
- current_ade20k_sem_seg_val
|
45 |
+
TRAIN:
|
46 |
+
- current_ade20k_sem_seg_train
|
47 |
+
GLOBAL:
|
48 |
+
HACK: 1.0
|
49 |
+
INPUT:
|
50 |
+
COLOR_AUG_SSD: true
|
51 |
+
CROP:
|
52 |
+
ENABLED: true
|
53 |
+
SINGLE_CATEGORY_MAX_AREA: 1.0
|
54 |
+
SIZE:
|
55 |
+
- 512
|
56 |
+
- 512
|
57 |
+
TYPE: absolute
|
58 |
+
DATASET_MAPPER_NAME: continual_semantic
|
59 |
+
FORMAT: RGB
|
60 |
+
IMAGE_SIZE: 1024
|
61 |
+
MASK_FORMAT: polygon
|
62 |
+
MAX_SCALE: 2.0
|
63 |
+
MAX_SIZE_TEST: 2048
|
64 |
+
MAX_SIZE_TRAIN: 2048
|
65 |
+
MIN_SCALE: 0.1
|
66 |
+
MIN_SIZE_TEST: 512
|
67 |
+
MIN_SIZE_TRAIN:
|
68 |
+
- 256
|
69 |
+
- 307
|
70 |
+
- 358
|
71 |
+
- 409
|
72 |
+
- 460
|
73 |
+
- 512
|
74 |
+
- 563
|
75 |
+
- 614
|
76 |
+
- 665
|
77 |
+
- 716
|
78 |
+
- 768
|
79 |
+
- 819
|
80 |
+
- 870
|
81 |
+
- 921
|
82 |
+
- 972
|
83 |
+
- 1024
|
84 |
+
MIN_SIZE_TRAIN_SAMPLING: choice
|
85 |
+
RANDOM_FLIP: horizontal
|
86 |
+
SIZE_DIVISIBILITY: 512
|
87 |
+
MODEL:
|
88 |
+
ANCHOR_GENERATOR:
|
89 |
+
ANGLES:
|
90 |
+
- - -90
|
91 |
+
- 0
|
92 |
+
- 90
|
93 |
+
ASPECT_RATIOS:
|
94 |
+
- - 0.5
|
95 |
+
- 1.0
|
96 |
+
- 2.0
|
97 |
+
NAME: DefaultAnchorGenerator
|
98 |
+
OFFSET: 0.0
|
99 |
+
SIZES:
|
100 |
+
- - 32
|
101 |
+
- 64
|
102 |
+
- 128
|
103 |
+
- 256
|
104 |
+
- 512
|
105 |
+
BACKBONE:
|
106 |
+
FREEZE_AT: 0
|
107 |
+
NAME: build_resnet_backbone
|
108 |
+
DEVICE: cuda
|
109 |
+
FPN:
|
110 |
+
FUSE_TYPE: sum
|
111 |
+
IN_FEATURES: []
|
112 |
+
NORM: ''
|
113 |
+
OUT_CHANNELS: 256
|
114 |
+
KEYPOINT_ON: false
|
115 |
+
LOAD_PROPOSALS: false
|
116 |
+
MASK_FORMER:
|
117 |
+
CLASS_WEIGHT: 2.0
|
118 |
+
CLIP_DIM: 512
|
119 |
+
DEC_LAYERS: 10
|
120 |
+
DEEP_SUPERVISION: true
|
121 |
+
DICE_WEIGHT: 5.0
|
122 |
+
DIM_FEEDFORWARD: 2048
|
123 |
+
DROPOUT: 0.0
|
124 |
+
ENC_LAYERS: 0
|
125 |
+
ENFORCE_INPUT_PROJ: false
|
126 |
+
HIDDEN_DIM: 256
|
127 |
+
IMPORTANCE_SAMPLE_RATIO: 0.75
|
128 |
+
MASK_WEIGHT: 5.0
|
129 |
+
NHEADS: 8
|
130 |
+
NO_OBJECT_WEIGHT: 0.1
|
131 |
+
NUM_OBJECT_QUERIES: 100
|
132 |
+
OVERSAMPLE_RATIO: 3.0
|
133 |
+
PRE_NORM: false
|
134 |
+
SIZE_DIVISIBILITY: 32
|
135 |
+
TEST:
|
136 |
+
INSTANCE_ON: false
|
137 |
+
OBJECT_MASK_THRESHOLD: 0.8
|
138 |
+
OVERLAP_THRESHOLD: 0.8
|
139 |
+
PANOPTIC_ON: false
|
140 |
+
SEMANTIC_ON: true
|
141 |
+
SEM_SEG_POSTPROCESSING_BEFORE_INFERENCE: false
|
142 |
+
TEXT_PATH: ''
|
143 |
+
TRAIN_NUM_POINTS: 12544
|
144 |
+
TRANSFORMER_DECODER_NAME: MultiScaleMaskedTransformerDecoder
|
145 |
+
TRANSFORMER_IN_FEATURE: multi_scale_pixel_decoder
|
146 |
+
USE_TEXT_EMBEDDING: false
|
147 |
+
MASK_ON: false
|
148 |
+
META_ARCHITECTURE: MaskFormer
|
149 |
+
PANOPTIC_FPN:
|
150 |
+
COMBINE:
|
151 |
+
ENABLED: true
|
152 |
+
INSTANCES_CONFIDENCE_THRESH: 0.5
|
153 |
+
OVERLAP_THRESH: 0.5
|
154 |
+
STUFF_AREA_LIMIT: 4096
|
155 |
+
INSTANCE_LOSS_WEIGHT: 1.0
|
156 |
+
PIXEL_MEAN:
|
157 |
+
- 123.675
|
158 |
+
- 116.28
|
159 |
+
- 103.53
|
160 |
+
PIXEL_STD:
|
161 |
+
- 58.395
|
162 |
+
- 57.12
|
163 |
+
- 57.375
|
164 |
+
PROPOSAL_GENERATOR:
|
165 |
+
MIN_SIZE: 0
|
166 |
+
NAME: RPN
|
167 |
+
RESNETS:
|
168 |
+
DEFORM_MODULATED: false
|
169 |
+
DEFORM_NUM_GROUPS: 1
|
170 |
+
DEFORM_ON_PER_STAGE:
|
171 |
+
- false
|
172 |
+
- false
|
173 |
+
- false
|
174 |
+
- false
|
175 |
+
DEPTH: 101
|
176 |
+
NORM: FrozenBN
|
177 |
+
NUM_GROUPS: 1
|
178 |
+
OUT_FEATURES:
|
179 |
+
- res2
|
180 |
+
- res3
|
181 |
+
- res4
|
182 |
+
- res5
|
183 |
+
RES2_OUT_CHANNELS: 256
|
184 |
+
RES4_DILATION: 1
|
185 |
+
RES5_DILATION: 1
|
186 |
+
RES5_MULTI_GRID:
|
187 |
+
- 1
|
188 |
+
- 1
|
189 |
+
- 1
|
190 |
+
STEM_OUT_CHANNELS: 64
|
191 |
+
STEM_TYPE: basic
|
192 |
+
STRIDE_IN_1X1: false
|
193 |
+
WIDTH_PER_GROUP: 64
|
194 |
+
RETINANET:
|
195 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
196 |
+
BBOX_REG_WEIGHTS: &id002
|
197 |
+
- 1.0
|
198 |
+
- 1.0
|
199 |
+
- 1.0
|
200 |
+
- 1.0
|
201 |
+
FOCAL_LOSS_ALPHA: 0.25
|
202 |
+
FOCAL_LOSS_GAMMA: 2.0
|
203 |
+
IN_FEATURES:
|
204 |
+
- p3
|
205 |
+
- p4
|
206 |
+
- p5
|
207 |
+
- p6
|
208 |
+
- p7
|
209 |
+
IOU_LABELS:
|
210 |
+
- 0
|
211 |
+
- -1
|
212 |
+
- 1
|
213 |
+
IOU_THRESHOLDS:
|
214 |
+
- 0.4
|
215 |
+
- 0.5
|
216 |
+
NMS_THRESH_TEST: 0.5
|
217 |
+
NORM: ''
|
218 |
+
NUM_CLASSES: 80
|
219 |
+
NUM_CONVS: 4
|
220 |
+
PRIOR_PROB: 0.01
|
221 |
+
SCORE_THRESH_TEST: 0.05
|
222 |
+
SMOOTH_L1_LOSS_BETA: 0.1
|
223 |
+
TOPK_CANDIDATES_TEST: 1000
|
224 |
+
ROI_BOX_CASCADE_HEAD:
|
225 |
+
BBOX_REG_WEIGHTS:
|
226 |
+
- &id001
|
227 |
+
- 10.0
|
228 |
+
- 10.0
|
229 |
+
- 5.0
|
230 |
+
- 5.0
|
231 |
+
- - 20.0
|
232 |
+
- 20.0
|
233 |
+
- 10.0
|
234 |
+
- 10.0
|
235 |
+
- - 30.0
|
236 |
+
- 30.0
|
237 |
+
- 15.0
|
238 |
+
- 15.0
|
239 |
+
IOUS:
|
240 |
+
- 0.5
|
241 |
+
- 0.6
|
242 |
+
- 0.7
|
243 |
+
ROI_BOX_HEAD:
|
244 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
245 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
246 |
+
BBOX_REG_WEIGHTS: *id001
|
247 |
+
CLS_AGNOSTIC_BBOX_REG: false
|
248 |
+
CONV_DIM: 256
|
249 |
+
FC_DIM: 1024
|
250 |
+
FED_LOSS_FREQ_WEIGHT_POWER: 0.5
|
251 |
+
FED_LOSS_NUM_CLASSES: 50
|
252 |
+
NAME: ''
|
253 |
+
NORM: ''
|
254 |
+
NUM_CONV: 0
|
255 |
+
NUM_FC: 0
|
256 |
+
POOLER_RESOLUTION: 14
|
257 |
+
POOLER_SAMPLING_RATIO: 0
|
258 |
+
POOLER_TYPE: ROIAlignV2
|
259 |
+
SMOOTH_L1_BETA: 0.0
|
260 |
+
TRAIN_ON_PRED_BOXES: false
|
261 |
+
USE_FED_LOSS: false
|
262 |
+
USE_SIGMOID_CE: false
|
263 |
+
ROI_HEADS:
|
264 |
+
BATCH_SIZE_PER_IMAGE: 512
|
265 |
+
IN_FEATURES:
|
266 |
+
- res4
|
267 |
+
IOU_LABELS:
|
268 |
+
- 0
|
269 |
+
- 1
|
270 |
+
IOU_THRESHOLDS:
|
271 |
+
- 0.5
|
272 |
+
NAME: Res5ROIHeads
|
273 |
+
NMS_THRESH_TEST: 0.5
|
274 |
+
NUM_CLASSES: 80
|
275 |
+
POSITIVE_FRACTION: 0.25
|
276 |
+
PROPOSAL_APPEND_GT: true
|
277 |
+
SCORE_THRESH_TEST: 0.05
|
278 |
+
ROI_KEYPOINT_HEAD:
|
279 |
+
CONV_DIMS:
|
280 |
+
- 512
|
281 |
+
- 512
|
282 |
+
- 512
|
283 |
+
- 512
|
284 |
+
- 512
|
285 |
+
- 512
|
286 |
+
- 512
|
287 |
+
- 512
|
288 |
+
LOSS_WEIGHT: 1.0
|
289 |
+
MIN_KEYPOINTS_PER_IMAGE: 1
|
290 |
+
NAME: KRCNNConvDeconvUpsampleHead
|
291 |
+
NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
|
292 |
+
NUM_KEYPOINTS: 17
|
293 |
+
POOLER_RESOLUTION: 14
|
294 |
+
POOLER_SAMPLING_RATIO: 0
|
295 |
+
POOLER_TYPE: ROIAlignV2
|
296 |
+
ROI_MASK_HEAD:
|
297 |
+
CLS_AGNOSTIC_MASK: false
|
298 |
+
CONV_DIM: 256
|
299 |
+
NAME: MaskRCNNConvUpsampleHead
|
300 |
+
NORM: ''
|
301 |
+
NUM_CONV: 0
|
302 |
+
POOLER_RESOLUTION: 14
|
303 |
+
POOLER_SAMPLING_RATIO: 0
|
304 |
+
POOLER_TYPE: ROIAlignV2
|
305 |
+
RPN:
|
306 |
+
BATCH_SIZE_PER_IMAGE: 256
|
307 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
308 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
309 |
+
BBOX_REG_WEIGHTS: *id002
|
310 |
+
BOUNDARY_THRESH: -1
|
311 |
+
CONV_DIMS:
|
312 |
+
- -1
|
313 |
+
HEAD_NAME: StandardRPNHead
|
314 |
+
IN_FEATURES:
|
315 |
+
- res4
|
316 |
+
IOU_LABELS:
|
317 |
+
- 0
|
318 |
+
- -1
|
319 |
+
- 1
|
320 |
+
IOU_THRESHOLDS:
|
321 |
+
- 0.3
|
322 |
+
- 0.7
|
323 |
+
LOSS_WEIGHT: 1.0
|
324 |
+
NMS_THRESH: 0.7
|
325 |
+
POSITIVE_FRACTION: 0.5
|
326 |
+
POST_NMS_TOPK_TEST: 1000
|
327 |
+
POST_NMS_TOPK_TRAIN: 2000
|
328 |
+
PRE_NMS_TOPK_TEST: 6000
|
329 |
+
PRE_NMS_TOPK_TRAIN: 12000
|
330 |
+
SMOOTH_L1_BETA: 0.0
|
331 |
+
SEM_SEG_HEAD:
|
332 |
+
ASPP_CHANNELS: 256
|
333 |
+
ASPP_DILATIONS:
|
334 |
+
- 6
|
335 |
+
- 12
|
336 |
+
- 18
|
337 |
+
ASPP_DROPOUT: 0.1
|
338 |
+
COMMON_STRIDE: 4
|
339 |
+
CONVS_DIM: 256
|
340 |
+
DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES:
|
341 |
+
- res3
|
342 |
+
- res4
|
343 |
+
- res5
|
344 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_HEADS: 8
|
345 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_POINTS: 4
|
346 |
+
IGNORE_VALUE: 255
|
347 |
+
IN_FEATURES:
|
348 |
+
- res2
|
349 |
+
- res3
|
350 |
+
- res4
|
351 |
+
- res5
|
352 |
+
LOSS_TYPE: hard_pixel_mining
|
353 |
+
LOSS_WEIGHT: 1.0
|
354 |
+
MASK_DIM: 256
|
355 |
+
NAME: MaskFormerHead
|
356 |
+
NORM: GN
|
357 |
+
NUM_CLASSES: 150
|
358 |
+
PIXEL_DECODER_NAME: MSDeformAttnPixelDecoder
|
359 |
+
PROJECT_CHANNELS:
|
360 |
+
- 48
|
361 |
+
PROJECT_FEATURES:
|
362 |
+
- res2
|
363 |
+
TRANSFORMER_ENC_LAYERS: 6
|
364 |
+
USE_DEPTHWISE_SEPARABLE_CONV: false
|
365 |
+
SWIN:
|
366 |
+
APE: false
|
367 |
+
ATTN_DROP_RATE: 0.0
|
368 |
+
DEPTHS:
|
369 |
+
- 2
|
370 |
+
- 2
|
371 |
+
- 6
|
372 |
+
- 2
|
373 |
+
DROP_PATH_RATE: 0.3
|
374 |
+
DROP_RATE: 0.0
|
375 |
+
EMBED_DIM: 96
|
376 |
+
MLP_RATIO: 4.0
|
377 |
+
NUM_HEADS:
|
378 |
+
- 3
|
379 |
+
- 6
|
380 |
+
- 12
|
381 |
+
- 24
|
382 |
+
OUT_FEATURES:
|
383 |
+
- res2
|
384 |
+
- res3
|
385 |
+
- res4
|
386 |
+
- res5
|
387 |
+
PATCH_NORM: true
|
388 |
+
PATCH_SIZE: 4
|
389 |
+
PRETRAIN_IMG_SIZE: 224
|
390 |
+
QKV_BIAS: true
|
391 |
+
QK_SCALE: null
|
392 |
+
USE_CHECKPOINT: false
|
393 |
+
WINDOW_SIZE: 7
|
394 |
+
WEIGHTS: ./output/newss/100-10_1_1+4_vq1/step5/model_final.pth
|
395 |
+
OUTPUT_DIR: ./output/newss/100-10_1_1+4_vq1/step6
|
396 |
+
SEED: -1
|
397 |
+
SOLVER:
|
398 |
+
AMP:
|
399 |
+
ENABLED: true
|
400 |
+
BACKBONE_MULTIPLIER: 0.1
|
401 |
+
BASE_LR: 5.0e-05
|
402 |
+
BASE_LR_END: 0.0
|
403 |
+
BIAS_LR_FACTOR: 1.0
|
404 |
+
CHECKPOINT_PERIOD: 10000
|
405 |
+
CLIP_GRADIENTS:
|
406 |
+
CLIP_TYPE: full_model
|
407 |
+
CLIP_VALUE: 0.01
|
408 |
+
ENABLED: true
|
409 |
+
NORM_TYPE: 2.0
|
410 |
+
GAMMA: 0.1
|
411 |
+
IMS_PER_BATCH: 8
|
412 |
+
LR_SCHEDULER_NAME: WarmupMultiStepLR
|
413 |
+
MAX_ITER: 10000
|
414 |
+
MOMENTUM: 0.9
|
415 |
+
NESTEROV: false
|
416 |
+
NUM_DECAYS: 3
|
417 |
+
OPTIMIZER: ADAMW
|
418 |
+
POLY_LR_CONSTANT_ENDING: 0.0
|
419 |
+
POLY_LR_POWER: 0.9
|
420 |
+
REFERENCE_WORLD_SIZE: 0
|
421 |
+
RESCALE_INTERVAL: false
|
422 |
+
STEPS:
|
423 |
+
- 8500
|
424 |
+
- 9500
|
425 |
+
WARMUP_FACTOR: 1.0
|
426 |
+
WARMUP_ITERS: 0
|
427 |
+
WARMUP_METHOD: linear
|
428 |
+
WEIGHT_DECAY: 0.05
|
429 |
+
WEIGHT_DECAY_BIAS: null
|
430 |
+
WEIGHT_DECAY_EMBED: 0.0
|
431 |
+
WEIGHT_DECAY_NORM: 0.0
|
432 |
+
TEST:
|
433 |
+
AUG:
|
434 |
+
ENABLED: false
|
435 |
+
FLIP: true
|
436 |
+
MAX_SIZE: 3584
|
437 |
+
MIN_SIZES:
|
438 |
+
- 256
|
439 |
+
- 384
|
440 |
+
- 512
|
441 |
+
- 640
|
442 |
+
- 768
|
443 |
+
- 896
|
444 |
+
DETECTIONS_PER_IMAGE: 100
|
445 |
+
EVAL_PERIOD: 5000
|
446 |
+
EXPECTED_RESULTS: []
|
447 |
+
KEYPOINT_OKS_SIGMAS: []
|
448 |
+
PRECISE_BN:
|
449 |
+
ENABLED: false
|
450 |
+
NUM_ITER: 200
|
451 |
+
VERSION: 2
|
452 |
+
VIS_PERIOD: 0
|
ss_100-10_42/step6/fake_query.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:804f309a9790c41453fe9ea4b0c596904abc30f3a3d4c9c8e0ddd3a49f8d6a0f
|
3 |
+
size 27746927
|
ss_100-10_42/step6/last_checkpoint
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
model_final.pth
|
ss_100-10_42/step6/metrics.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
ss_100-10_42/step6/model_final.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:eeb45ba20d09d464d06edad795fa833e8d69bb6812c902551c3c81d10adbb102
|
3 |
+
size 762308965
|
ss_100-10_42/step6/psd_distribution.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
[2251.0, 519.0, 477.0, 1820.0, 348.0, 1064.0, 235.0, 380.0, 914.0, 120.0, 779.0, 275.0, 679.0, 38.0, 496.0, 1172.0, 53.0, 598.0, 492.0, 944.0, 235.0, 21.0, 694.0, 259.0, 337.0, 2.0, 37.0, 246.0, 186.0, 2.0, 152.0, 59.0, 55.0, 155.0, 20.0, 33.0, 757.0, 85.0, 61.0, 321.0, 23.0, 605.0, 111.0, 204.0, 71.0, 115.0, 6.0, 338.0, 22.0, 118.0, 65.0, 8.0, 16.0, 49.0, 0.0, 38.0, 37.0, 208.0, 29.0, 22.0, 4.0, 15.0, 51.0, 40.0, 135.0, 86.0, 311.0, 468.0, 2.0, 43.0, 224.0, 163.0, 18.0, 72.0, 82.0, 65.0, 24.0, 28.0, 19.0, 0.0, 10.0, 116.0, 384.0, 14.0, 24.0, 138.0, 57.0, 75.0, 8.0, 138.0, 8.0, 0.0, 15.0, 78.0, 0.0, 18.0, 5.0, 44.0, 510.0, 9.0, 263.0, 39.0, 207.0, 24.0, 20.0, 29.0, 16.0, 43.0, 770.0, 19.0, 323.0, 0.0, 371.0, 2.0, 7.0, 264.0, 94.0, 9.0, 57.0, 29.0, 25.0, 101.0, 6.0, 47.0, 113.0, 416.0, 1.0, 63.0, 34.0, 74.0, 19.0, 22.0, 29.0, 56.0, 142.0, 298.0, 20.0, 95.0, 155.0, 50.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
|
ss_100-50_48/step2/config.yaml
ADDED
@@ -0,0 +1,452 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
CONT:
|
2 |
+
ADD_POS: false
|
3 |
+
BASE_CLS: 100
|
4 |
+
COLLECT_QUERY_MODE: false
|
5 |
+
COMBINE_PSDLABEL: true
|
6 |
+
CUMULATIVE_PSDNUM: false
|
7 |
+
DISTRIBUTION_ALPHA: 0.25
|
8 |
+
FILTER_KD: false
|
9 |
+
FREEZE_LABEL: false
|
10 |
+
INC_CLS: 50
|
11 |
+
KD_DECODER: false
|
12 |
+
KD_TEMPERATURE: 1.0
|
13 |
+
KD_TEMPERATURE2: 10.0
|
14 |
+
KD_TYPE: kl
|
15 |
+
KL_ALL: true
|
16 |
+
KL_WEIGHT: 2.0
|
17 |
+
LIB_SIZE: 80
|
18 |
+
MED_TOKENS_WEIGHT: 5.0
|
19 |
+
MEMORY: false
|
20 |
+
OLD_WEIGHTS: output/newss/100-50_1_1+0_vq1/step1/model_final.pth
|
21 |
+
PSD_LABEL_THRESHOLD: 0.0
|
22 |
+
PSD_OVERLAP_THRESHOLD: 0.8
|
23 |
+
SETTING: overlapped
|
24 |
+
TASK: 2
|
25 |
+
TOT_CLS: 150
|
26 |
+
VQ_NUMBER: 1
|
27 |
+
VQ_STORE: false
|
28 |
+
WEIGHTED_SAMPLE: true
|
29 |
+
WEIGHTS: null
|
30 |
+
CUDNN_BENCHMARK: false
|
31 |
+
DATALOADER:
|
32 |
+
ASPECT_RATIO_GROUPING: true
|
33 |
+
FILTER_EMPTY_ANNOTATIONS: true
|
34 |
+
NUM_WORKERS: 2
|
35 |
+
REPEAT_SQRT: true
|
36 |
+
REPEAT_THRESHOLD: 0.0
|
37 |
+
SAMPLER_TRAIN: TrainingSampler
|
38 |
+
DATASETS:
|
39 |
+
PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
|
40 |
+
PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
|
41 |
+
PROPOSAL_FILES_TEST: []
|
42 |
+
PROPOSAL_FILES_TRAIN: []
|
43 |
+
TEST:
|
44 |
+
- current_ade20k_sem_seg_val
|
45 |
+
TRAIN:
|
46 |
+
- current_ade20k_sem_seg_train
|
47 |
+
GLOBAL:
|
48 |
+
HACK: 1.0
|
49 |
+
INPUT:
|
50 |
+
COLOR_AUG_SSD: true
|
51 |
+
CROP:
|
52 |
+
ENABLED: true
|
53 |
+
SINGLE_CATEGORY_MAX_AREA: 1.0
|
54 |
+
SIZE:
|
55 |
+
- 512
|
56 |
+
- 512
|
57 |
+
TYPE: absolute
|
58 |
+
DATASET_MAPPER_NAME: continual_semantic
|
59 |
+
FORMAT: RGB
|
60 |
+
IMAGE_SIZE: 1024
|
61 |
+
MASK_FORMAT: polygon
|
62 |
+
MAX_SCALE: 2.0
|
63 |
+
MAX_SIZE_TEST: 2048
|
64 |
+
MAX_SIZE_TRAIN: 2048
|
65 |
+
MIN_SCALE: 0.1
|
66 |
+
MIN_SIZE_TEST: 512
|
67 |
+
MIN_SIZE_TRAIN:
|
68 |
+
- 256
|
69 |
+
- 307
|
70 |
+
- 358
|
71 |
+
- 409
|
72 |
+
- 460
|
73 |
+
- 512
|
74 |
+
- 563
|
75 |
+
- 614
|
76 |
+
- 665
|
77 |
+
- 716
|
78 |
+
- 768
|
79 |
+
- 819
|
80 |
+
- 870
|
81 |
+
- 921
|
82 |
+
- 972
|
83 |
+
- 1024
|
84 |
+
MIN_SIZE_TRAIN_SAMPLING: choice
|
85 |
+
RANDOM_FLIP: horizontal
|
86 |
+
SIZE_DIVISIBILITY: 512
|
87 |
+
MODEL:
|
88 |
+
ANCHOR_GENERATOR:
|
89 |
+
ANGLES:
|
90 |
+
- - -90
|
91 |
+
- 0
|
92 |
+
- 90
|
93 |
+
ASPECT_RATIOS:
|
94 |
+
- - 0.5
|
95 |
+
- 1.0
|
96 |
+
- 2.0
|
97 |
+
NAME: DefaultAnchorGenerator
|
98 |
+
OFFSET: 0.0
|
99 |
+
SIZES:
|
100 |
+
- - 32
|
101 |
+
- 64
|
102 |
+
- 128
|
103 |
+
- 256
|
104 |
+
- 512
|
105 |
+
BACKBONE:
|
106 |
+
FREEZE_AT: 0
|
107 |
+
NAME: build_resnet_backbone
|
108 |
+
DEVICE: cuda
|
109 |
+
FPN:
|
110 |
+
FUSE_TYPE: sum
|
111 |
+
IN_FEATURES: []
|
112 |
+
NORM: ''
|
113 |
+
OUT_CHANNELS: 256
|
114 |
+
KEYPOINT_ON: false
|
115 |
+
LOAD_PROPOSALS: false
|
116 |
+
MASK_FORMER:
|
117 |
+
CLASS_WEIGHT: 2.0
|
118 |
+
CLIP_DIM: 512
|
119 |
+
DEC_LAYERS: 10
|
120 |
+
DEEP_SUPERVISION: true
|
121 |
+
DICE_WEIGHT: 5.0
|
122 |
+
DIM_FEEDFORWARD: 2048
|
123 |
+
DROPOUT: 0.0
|
124 |
+
ENC_LAYERS: 0
|
125 |
+
ENFORCE_INPUT_PROJ: false
|
126 |
+
HIDDEN_DIM: 256
|
127 |
+
IMPORTANCE_SAMPLE_RATIO: 0.75
|
128 |
+
MASK_WEIGHT: 5.0
|
129 |
+
NHEADS: 8
|
130 |
+
NO_OBJECT_WEIGHT: 0.1
|
131 |
+
NUM_OBJECT_QUERIES: 100
|
132 |
+
OVERSAMPLE_RATIO: 3.0
|
133 |
+
PRE_NORM: false
|
134 |
+
SIZE_DIVISIBILITY: 32
|
135 |
+
TEST:
|
136 |
+
INSTANCE_ON: false
|
137 |
+
OBJECT_MASK_THRESHOLD: 0.8
|
138 |
+
OVERLAP_THRESHOLD: 0.8
|
139 |
+
PANOPTIC_ON: false
|
140 |
+
SEMANTIC_ON: true
|
141 |
+
SEM_SEG_POSTPROCESSING_BEFORE_INFERENCE: false
|
142 |
+
TEXT_PATH: ''
|
143 |
+
TRAIN_NUM_POINTS: 12544
|
144 |
+
TRANSFORMER_DECODER_NAME: MultiScaleMaskedTransformerDecoder
|
145 |
+
TRANSFORMER_IN_FEATURE: multi_scale_pixel_decoder
|
146 |
+
USE_TEXT_EMBEDDING: false
|
147 |
+
MASK_ON: false
|
148 |
+
META_ARCHITECTURE: MaskFormer
|
149 |
+
PANOPTIC_FPN:
|
150 |
+
COMBINE:
|
151 |
+
ENABLED: true
|
152 |
+
INSTANCES_CONFIDENCE_THRESH: 0.5
|
153 |
+
OVERLAP_THRESH: 0.5
|
154 |
+
STUFF_AREA_LIMIT: 4096
|
155 |
+
INSTANCE_LOSS_WEIGHT: 1.0
|
156 |
+
PIXEL_MEAN:
|
157 |
+
- 123.675
|
158 |
+
- 116.28
|
159 |
+
- 103.53
|
160 |
+
PIXEL_STD:
|
161 |
+
- 58.395
|
162 |
+
- 57.12
|
163 |
+
- 57.375
|
164 |
+
PROPOSAL_GENERATOR:
|
165 |
+
MIN_SIZE: 0
|
166 |
+
NAME: RPN
|
167 |
+
RESNETS:
|
168 |
+
DEFORM_MODULATED: false
|
169 |
+
DEFORM_NUM_GROUPS: 1
|
170 |
+
DEFORM_ON_PER_STAGE:
|
171 |
+
- false
|
172 |
+
- false
|
173 |
+
- false
|
174 |
+
- false
|
175 |
+
DEPTH: 101
|
176 |
+
NORM: FrozenBN
|
177 |
+
NUM_GROUPS: 1
|
178 |
+
OUT_FEATURES:
|
179 |
+
- res2
|
180 |
+
- res3
|
181 |
+
- res4
|
182 |
+
- res5
|
183 |
+
RES2_OUT_CHANNELS: 256
|
184 |
+
RES4_DILATION: 1
|
185 |
+
RES5_DILATION: 1
|
186 |
+
RES5_MULTI_GRID:
|
187 |
+
- 1
|
188 |
+
- 1
|
189 |
+
- 1
|
190 |
+
STEM_OUT_CHANNELS: 64
|
191 |
+
STEM_TYPE: basic
|
192 |
+
STRIDE_IN_1X1: false
|
193 |
+
WIDTH_PER_GROUP: 64
|
194 |
+
RETINANET:
|
195 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
196 |
+
BBOX_REG_WEIGHTS: &id002
|
197 |
+
- 1.0
|
198 |
+
- 1.0
|
199 |
+
- 1.0
|
200 |
+
- 1.0
|
201 |
+
FOCAL_LOSS_ALPHA: 0.25
|
202 |
+
FOCAL_LOSS_GAMMA: 2.0
|
203 |
+
IN_FEATURES:
|
204 |
+
- p3
|
205 |
+
- p4
|
206 |
+
- p5
|
207 |
+
- p6
|
208 |
+
- p7
|
209 |
+
IOU_LABELS:
|
210 |
+
- 0
|
211 |
+
- -1
|
212 |
+
- 1
|
213 |
+
IOU_THRESHOLDS:
|
214 |
+
- 0.4
|
215 |
+
- 0.5
|
216 |
+
NMS_THRESH_TEST: 0.5
|
217 |
+
NORM: ''
|
218 |
+
NUM_CLASSES: 80
|
219 |
+
NUM_CONVS: 4
|
220 |
+
PRIOR_PROB: 0.01
|
221 |
+
SCORE_THRESH_TEST: 0.05
|
222 |
+
SMOOTH_L1_LOSS_BETA: 0.1
|
223 |
+
TOPK_CANDIDATES_TEST: 1000
|
224 |
+
ROI_BOX_CASCADE_HEAD:
|
225 |
+
BBOX_REG_WEIGHTS:
|
226 |
+
- &id001
|
227 |
+
- 10.0
|
228 |
+
- 10.0
|
229 |
+
- 5.0
|
230 |
+
- 5.0
|
231 |
+
- - 20.0
|
232 |
+
- 20.0
|
233 |
+
- 10.0
|
234 |
+
- 10.0
|
235 |
+
- - 30.0
|
236 |
+
- 30.0
|
237 |
+
- 15.0
|
238 |
+
- 15.0
|
239 |
+
IOUS:
|
240 |
+
- 0.5
|
241 |
+
- 0.6
|
242 |
+
- 0.7
|
243 |
+
ROI_BOX_HEAD:
|
244 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
245 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
246 |
+
BBOX_REG_WEIGHTS: *id001
|
247 |
+
CLS_AGNOSTIC_BBOX_REG: false
|
248 |
+
CONV_DIM: 256
|
249 |
+
FC_DIM: 1024
|
250 |
+
FED_LOSS_FREQ_WEIGHT_POWER: 0.5
|
251 |
+
FED_LOSS_NUM_CLASSES: 50
|
252 |
+
NAME: ''
|
253 |
+
NORM: ''
|
254 |
+
NUM_CONV: 0
|
255 |
+
NUM_FC: 0
|
256 |
+
POOLER_RESOLUTION: 14
|
257 |
+
POOLER_SAMPLING_RATIO: 0
|
258 |
+
POOLER_TYPE: ROIAlignV2
|
259 |
+
SMOOTH_L1_BETA: 0.0
|
260 |
+
TRAIN_ON_PRED_BOXES: false
|
261 |
+
USE_FED_LOSS: false
|
262 |
+
USE_SIGMOID_CE: false
|
263 |
+
ROI_HEADS:
|
264 |
+
BATCH_SIZE_PER_IMAGE: 512
|
265 |
+
IN_FEATURES:
|
266 |
+
- res4
|
267 |
+
IOU_LABELS:
|
268 |
+
- 0
|
269 |
+
- 1
|
270 |
+
IOU_THRESHOLDS:
|
271 |
+
- 0.5
|
272 |
+
NAME: Res5ROIHeads
|
273 |
+
NMS_THRESH_TEST: 0.5
|
274 |
+
NUM_CLASSES: 80
|
275 |
+
POSITIVE_FRACTION: 0.25
|
276 |
+
PROPOSAL_APPEND_GT: true
|
277 |
+
SCORE_THRESH_TEST: 0.05
|
278 |
+
ROI_KEYPOINT_HEAD:
|
279 |
+
CONV_DIMS:
|
280 |
+
- 512
|
281 |
+
- 512
|
282 |
+
- 512
|
283 |
+
- 512
|
284 |
+
- 512
|
285 |
+
- 512
|
286 |
+
- 512
|
287 |
+
- 512
|
288 |
+
LOSS_WEIGHT: 1.0
|
289 |
+
MIN_KEYPOINTS_PER_IMAGE: 1
|
290 |
+
NAME: KRCNNConvDeconvUpsampleHead
|
291 |
+
NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
|
292 |
+
NUM_KEYPOINTS: 17
|
293 |
+
POOLER_RESOLUTION: 14
|
294 |
+
POOLER_SAMPLING_RATIO: 0
|
295 |
+
POOLER_TYPE: ROIAlignV2
|
296 |
+
ROI_MASK_HEAD:
|
297 |
+
CLS_AGNOSTIC_MASK: false
|
298 |
+
CONV_DIM: 256
|
299 |
+
NAME: MaskRCNNConvUpsampleHead
|
300 |
+
NORM: ''
|
301 |
+
NUM_CONV: 0
|
302 |
+
POOLER_RESOLUTION: 14
|
303 |
+
POOLER_SAMPLING_RATIO: 0
|
304 |
+
POOLER_TYPE: ROIAlignV2
|
305 |
+
RPN:
|
306 |
+
BATCH_SIZE_PER_IMAGE: 256
|
307 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
308 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
309 |
+
BBOX_REG_WEIGHTS: *id002
|
310 |
+
BOUNDARY_THRESH: -1
|
311 |
+
CONV_DIMS:
|
312 |
+
- -1
|
313 |
+
HEAD_NAME: StandardRPNHead
|
314 |
+
IN_FEATURES:
|
315 |
+
- res4
|
316 |
+
IOU_LABELS:
|
317 |
+
- 0
|
318 |
+
- -1
|
319 |
+
- 1
|
320 |
+
IOU_THRESHOLDS:
|
321 |
+
- 0.3
|
322 |
+
- 0.7
|
323 |
+
LOSS_WEIGHT: 1.0
|
324 |
+
NMS_THRESH: 0.7
|
325 |
+
POSITIVE_FRACTION: 0.5
|
326 |
+
POST_NMS_TOPK_TEST: 1000
|
327 |
+
POST_NMS_TOPK_TRAIN: 2000
|
328 |
+
PRE_NMS_TOPK_TEST: 6000
|
329 |
+
PRE_NMS_TOPK_TRAIN: 12000
|
330 |
+
SMOOTH_L1_BETA: 0.0
|
331 |
+
SEM_SEG_HEAD:
|
332 |
+
ASPP_CHANNELS: 256
|
333 |
+
ASPP_DILATIONS:
|
334 |
+
- 6
|
335 |
+
- 12
|
336 |
+
- 18
|
337 |
+
ASPP_DROPOUT: 0.1
|
338 |
+
COMMON_STRIDE: 4
|
339 |
+
CONVS_DIM: 256
|
340 |
+
DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES:
|
341 |
+
- res3
|
342 |
+
- res4
|
343 |
+
- res5
|
344 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_HEADS: 8
|
345 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_POINTS: 4
|
346 |
+
IGNORE_VALUE: 255
|
347 |
+
IN_FEATURES:
|
348 |
+
- res2
|
349 |
+
- res3
|
350 |
+
- res4
|
351 |
+
- res5
|
352 |
+
LOSS_TYPE: hard_pixel_mining
|
353 |
+
LOSS_WEIGHT: 1.0
|
354 |
+
MASK_DIM: 256
|
355 |
+
NAME: MaskFormerHead
|
356 |
+
NORM: GN
|
357 |
+
NUM_CLASSES: 150
|
358 |
+
PIXEL_DECODER_NAME: MSDeformAttnPixelDecoder
|
359 |
+
PROJECT_CHANNELS:
|
360 |
+
- 48
|
361 |
+
PROJECT_FEATURES:
|
362 |
+
- res2
|
363 |
+
TRANSFORMER_ENC_LAYERS: 6
|
364 |
+
USE_DEPTHWISE_SEPARABLE_CONV: false
|
365 |
+
SWIN:
|
366 |
+
APE: false
|
367 |
+
ATTN_DROP_RATE: 0.0
|
368 |
+
DEPTHS:
|
369 |
+
- 2
|
370 |
+
- 2
|
371 |
+
- 6
|
372 |
+
- 2
|
373 |
+
DROP_PATH_RATE: 0.3
|
374 |
+
DROP_RATE: 0.0
|
375 |
+
EMBED_DIM: 96
|
376 |
+
MLP_RATIO: 4.0
|
377 |
+
NUM_HEADS:
|
378 |
+
- 3
|
379 |
+
- 6
|
380 |
+
- 12
|
381 |
+
- 24
|
382 |
+
OUT_FEATURES:
|
383 |
+
- res2
|
384 |
+
- res3
|
385 |
+
- res4
|
386 |
+
- res5
|
387 |
+
PATCH_NORM: true
|
388 |
+
PATCH_SIZE: 4
|
389 |
+
PRETRAIN_IMG_SIZE: 224
|
390 |
+
QKV_BIAS: true
|
391 |
+
QK_SCALE: null
|
392 |
+
USE_CHECKPOINT: false
|
393 |
+
WINDOW_SIZE: 7
|
394 |
+
WEIGHTS: output/newss/100-50_1_1+0_vq1/step1/model_final.pth
|
395 |
+
OUTPUT_DIR: output/newss/100-50_1_1+0_vq1/step2
|
396 |
+
SEED: -1
|
397 |
+
SOLVER:
|
398 |
+
AMP:
|
399 |
+
ENABLED: true
|
400 |
+
BACKBONE_MULTIPLIER: 0.1
|
401 |
+
BASE_LR: 5.0e-05
|
402 |
+
BASE_LR_END: 0.0
|
403 |
+
BIAS_LR_FACTOR: 1.0
|
404 |
+
CHECKPOINT_PERIOD: 10000
|
405 |
+
CLIP_GRADIENTS:
|
406 |
+
CLIP_TYPE: full_model
|
407 |
+
CLIP_VALUE: 0.01
|
408 |
+
ENABLED: true
|
409 |
+
NORM_TYPE: 2.0
|
410 |
+
GAMMA: 0.1
|
411 |
+
IMS_PER_BATCH: 8
|
412 |
+
LR_SCHEDULER_NAME: WarmupMultiStepLR
|
413 |
+
MAX_ITER: 50000
|
414 |
+
MOMENTUM: 0.9
|
415 |
+
NESTEROV: false
|
416 |
+
NUM_DECAYS: 3
|
417 |
+
OPTIMIZER: ADAMW
|
418 |
+
POLY_LR_CONSTANT_ENDING: 0.0
|
419 |
+
POLY_LR_POWER: 0.9
|
420 |
+
REFERENCE_WORLD_SIZE: 0
|
421 |
+
RESCALE_INTERVAL: false
|
422 |
+
STEPS:
|
423 |
+
- 42500
|
424 |
+
- 47500
|
425 |
+
WARMUP_FACTOR: 1.0
|
426 |
+
WARMUP_ITERS: 0
|
427 |
+
WARMUP_METHOD: linear
|
428 |
+
WEIGHT_DECAY: 0.05
|
429 |
+
WEIGHT_DECAY_BIAS: null
|
430 |
+
WEIGHT_DECAY_EMBED: 0.0
|
431 |
+
WEIGHT_DECAY_NORM: 0.0
|
432 |
+
TEST:
|
433 |
+
AUG:
|
434 |
+
ENABLED: false
|
435 |
+
FLIP: true
|
436 |
+
MAX_SIZE: 3584
|
437 |
+
MIN_SIZES:
|
438 |
+
- 256
|
439 |
+
- 384
|
440 |
+
- 512
|
441 |
+
- 640
|
442 |
+
- 768
|
443 |
+
- 896
|
444 |
+
DETECTIONS_PER_IMAGE: 100
|
445 |
+
EVAL_PERIOD: 5000
|
446 |
+
EXPECTED_RESULTS: []
|
447 |
+
KEYPOINT_OKS_SIGMAS: []
|
448 |
+
PRECISE_BN:
|
449 |
+
ENABLED: false
|
450 |
+
NUM_ITER: 200
|
451 |
+
VERSION: 2
|
452 |
+
VIS_PERIOD: 0
|
ss_100-50_48/step2/fake_query.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2235c540abf79c13040f995e22a70836f9f49a0928521cb309b8fc14fc10fc42
|
3 |
+
size 27746927
|
ss_100-50_48/step2/last_checkpoint
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
model_final.pth
|
ss_100-50_48/step2/metrics.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
ss_100-50_48/step2/model_final.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f86dfab2e29fa2b78147540cdbd0a21666d733ccbbfccd0a7494ed96ce487039
|
3 |
+
size 762308965
|
ss_100-50_48/step2/psd_distribution.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
[9398.0, 3769.0, 3245.0, 6995.0, 2464.0, 4558.0, 2298.0, 1689.0, 3649.0, 834.0, 2643.0, 2066.0, 2951.0, 368.0, 1623.0, 3682.0, 362.0, 2187.0, 1861.0, 2633.0, 1962.0, 260.0, 2823.0, 1099.0, 772.0, 108.0, 131.0, 930.0, 914.0, 60.0, 813.0, 310.0, 344.0, 440.0, 127.0, 252.0, 2328.0, 317.0, 269.0, 1326.0, 203.0, 1122.0, 385.0, 939.0, 474.0, 311.0, 25.0, 1165.0, 73.0, 506.0, 367.0, 38.0, 81.0, 225.0, 27.0, 111.0, 197.0, 634.0, 87.0, 151.0, 28.0, 56.0, 173.0, 156.0, 755.0, 286.0, 1095.0, 832.0, 10.0, 239.0, 408.0, 651.0, 104.0, 226.0, 216.0, 228.0, 84.0, 99.0, 58.0, 9.0, 108.0, 410.0, 909.0, 190.0, 64.0, 484.0, 280.0, 230.0, 30.0, 479.0, 36.0, 2.0, 113.0, 290.0, 8.0, 43.0, 22.0, 284.0, 736.0, 64.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
|
ss_100-5_38/step11/config.yaml
ADDED
@@ -0,0 +1,452 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
CONT:
|
2 |
+
ADD_POS: false
|
3 |
+
BASE_CLS: 100
|
4 |
+
COLLECT_QUERY_MODE: false
|
5 |
+
COMBINE_PSDLABEL: true
|
6 |
+
CUMULATIVE_PSDNUM: false
|
7 |
+
DISTRIBUTION_ALPHA: 0.5
|
8 |
+
FILTER_KD: false
|
9 |
+
FREEZE_LABEL: false
|
10 |
+
INC_CLS: 5
|
11 |
+
KD_DECODER: true
|
12 |
+
KD_TEMPERATURE: 1.0
|
13 |
+
KD_TEMPERATURE2: 2.0
|
14 |
+
KD_TYPE: kl
|
15 |
+
KL_ALL: true
|
16 |
+
KL_WEIGHT: 2.0
|
17 |
+
LIB_SIZE: 80
|
18 |
+
MED_TOKENS_WEIGHT: 5.0
|
19 |
+
MEMORY: false
|
20 |
+
OLD_WEIGHTS: ./output/newss/100-5_1_1+2_vq3/step10/model_final.pth
|
21 |
+
PSD_LABEL_THRESHOLD: 0.0
|
22 |
+
PSD_OVERLAP_THRESHOLD: 0.8
|
23 |
+
SETTING: overlapped
|
24 |
+
TASK: 11
|
25 |
+
TOT_CLS: 150
|
26 |
+
VQ_NUMBER: 3
|
27 |
+
VQ_STORE: false
|
28 |
+
WEIGHTED_SAMPLE: true
|
29 |
+
WEIGHTS: null
|
30 |
+
CUDNN_BENCHMARK: false
|
31 |
+
DATALOADER:
|
32 |
+
ASPECT_RATIO_GROUPING: true
|
33 |
+
FILTER_EMPTY_ANNOTATIONS: true
|
34 |
+
NUM_WORKERS: 2
|
35 |
+
REPEAT_SQRT: true
|
36 |
+
REPEAT_THRESHOLD: 0.0
|
37 |
+
SAMPLER_TRAIN: TrainingSampler
|
38 |
+
DATASETS:
|
39 |
+
PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
|
40 |
+
PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
|
41 |
+
PROPOSAL_FILES_TEST: []
|
42 |
+
PROPOSAL_FILES_TRAIN: []
|
43 |
+
TEST:
|
44 |
+
- current_ade20k_sem_seg_val
|
45 |
+
TRAIN:
|
46 |
+
- current_ade20k_sem_seg_train
|
47 |
+
GLOBAL:
|
48 |
+
HACK: 1.0
|
49 |
+
INPUT:
|
50 |
+
COLOR_AUG_SSD: true
|
51 |
+
CROP:
|
52 |
+
ENABLED: true
|
53 |
+
SINGLE_CATEGORY_MAX_AREA: 1.0
|
54 |
+
SIZE:
|
55 |
+
- 512
|
56 |
+
- 512
|
57 |
+
TYPE: absolute
|
58 |
+
DATASET_MAPPER_NAME: continual_semantic
|
59 |
+
FORMAT: RGB
|
60 |
+
IMAGE_SIZE: 1024
|
61 |
+
MASK_FORMAT: polygon
|
62 |
+
MAX_SCALE: 2.0
|
63 |
+
MAX_SIZE_TEST: 2048
|
64 |
+
MAX_SIZE_TRAIN: 2048
|
65 |
+
MIN_SCALE: 0.1
|
66 |
+
MIN_SIZE_TEST: 512
|
67 |
+
MIN_SIZE_TRAIN:
|
68 |
+
- 256
|
69 |
+
- 307
|
70 |
+
- 358
|
71 |
+
- 409
|
72 |
+
- 460
|
73 |
+
- 512
|
74 |
+
- 563
|
75 |
+
- 614
|
76 |
+
- 665
|
77 |
+
- 716
|
78 |
+
- 768
|
79 |
+
- 819
|
80 |
+
- 870
|
81 |
+
- 921
|
82 |
+
- 972
|
83 |
+
- 1024
|
84 |
+
MIN_SIZE_TRAIN_SAMPLING: choice
|
85 |
+
RANDOM_FLIP: horizontal
|
86 |
+
SIZE_DIVISIBILITY: 512
|
87 |
+
MODEL:
|
88 |
+
ANCHOR_GENERATOR:
|
89 |
+
ANGLES:
|
90 |
+
- - -90
|
91 |
+
- 0
|
92 |
+
- 90
|
93 |
+
ASPECT_RATIOS:
|
94 |
+
- - 0.5
|
95 |
+
- 1.0
|
96 |
+
- 2.0
|
97 |
+
NAME: DefaultAnchorGenerator
|
98 |
+
OFFSET: 0.0
|
99 |
+
SIZES:
|
100 |
+
- - 32
|
101 |
+
- 64
|
102 |
+
- 128
|
103 |
+
- 256
|
104 |
+
- 512
|
105 |
+
BACKBONE:
|
106 |
+
FREEZE_AT: 0
|
107 |
+
NAME: build_resnet_backbone
|
108 |
+
DEVICE: cuda
|
109 |
+
FPN:
|
110 |
+
FUSE_TYPE: sum
|
111 |
+
IN_FEATURES: []
|
112 |
+
NORM: ''
|
113 |
+
OUT_CHANNELS: 256
|
114 |
+
KEYPOINT_ON: false
|
115 |
+
LOAD_PROPOSALS: false
|
116 |
+
MASK_FORMER:
|
117 |
+
CLASS_WEIGHT: 2.0
|
118 |
+
CLIP_DIM: 512
|
119 |
+
DEC_LAYERS: 10
|
120 |
+
DEEP_SUPERVISION: true
|
121 |
+
DICE_WEIGHT: 5.0
|
122 |
+
DIM_FEEDFORWARD: 2048
|
123 |
+
DROPOUT: 0.0
|
124 |
+
ENC_LAYERS: 0
|
125 |
+
ENFORCE_INPUT_PROJ: false
|
126 |
+
HIDDEN_DIM: 256
|
127 |
+
IMPORTANCE_SAMPLE_RATIO: 0.75
|
128 |
+
MASK_WEIGHT: 5.0
|
129 |
+
NHEADS: 8
|
130 |
+
NO_OBJECT_WEIGHT: 0.1
|
131 |
+
NUM_OBJECT_QUERIES: 100
|
132 |
+
OVERSAMPLE_RATIO: 3.0
|
133 |
+
PRE_NORM: false
|
134 |
+
SIZE_DIVISIBILITY: 32
|
135 |
+
TEST:
|
136 |
+
INSTANCE_ON: false
|
137 |
+
OBJECT_MASK_THRESHOLD: 0.8
|
138 |
+
OVERLAP_THRESHOLD: 0.8
|
139 |
+
PANOPTIC_ON: false
|
140 |
+
SEMANTIC_ON: true
|
141 |
+
SEM_SEG_POSTPROCESSING_BEFORE_INFERENCE: false
|
142 |
+
TEXT_PATH: ''
|
143 |
+
TRAIN_NUM_POINTS: 12544
|
144 |
+
TRANSFORMER_DECODER_NAME: MultiScaleMaskedTransformerDecoder
|
145 |
+
TRANSFORMER_IN_FEATURE: multi_scale_pixel_decoder
|
146 |
+
USE_TEXT_EMBEDDING: false
|
147 |
+
MASK_ON: false
|
148 |
+
META_ARCHITECTURE: MaskFormer
|
149 |
+
PANOPTIC_FPN:
|
150 |
+
COMBINE:
|
151 |
+
ENABLED: true
|
152 |
+
INSTANCES_CONFIDENCE_THRESH: 0.5
|
153 |
+
OVERLAP_THRESH: 0.5
|
154 |
+
STUFF_AREA_LIMIT: 4096
|
155 |
+
INSTANCE_LOSS_WEIGHT: 1.0
|
156 |
+
PIXEL_MEAN:
|
157 |
+
- 123.675
|
158 |
+
- 116.28
|
159 |
+
- 103.53
|
160 |
+
PIXEL_STD:
|
161 |
+
- 58.395
|
162 |
+
- 57.12
|
163 |
+
- 57.375
|
164 |
+
PROPOSAL_GENERATOR:
|
165 |
+
MIN_SIZE: 0
|
166 |
+
NAME: RPN
|
167 |
+
RESNETS:
|
168 |
+
DEFORM_MODULATED: false
|
169 |
+
DEFORM_NUM_GROUPS: 1
|
170 |
+
DEFORM_ON_PER_STAGE:
|
171 |
+
- false
|
172 |
+
- false
|
173 |
+
- false
|
174 |
+
- false
|
175 |
+
DEPTH: 101
|
176 |
+
NORM: FrozenBN
|
177 |
+
NUM_GROUPS: 1
|
178 |
+
OUT_FEATURES:
|
179 |
+
- res2
|
180 |
+
- res3
|
181 |
+
- res4
|
182 |
+
- res5
|
183 |
+
RES2_OUT_CHANNELS: 256
|
184 |
+
RES4_DILATION: 1
|
185 |
+
RES5_DILATION: 1
|
186 |
+
RES5_MULTI_GRID:
|
187 |
+
- 1
|
188 |
+
- 1
|
189 |
+
- 1
|
190 |
+
STEM_OUT_CHANNELS: 64
|
191 |
+
STEM_TYPE: basic
|
192 |
+
STRIDE_IN_1X1: false
|
193 |
+
WIDTH_PER_GROUP: 64
|
194 |
+
RETINANET:
|
195 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
196 |
+
BBOX_REG_WEIGHTS: &id002
|
197 |
+
- 1.0
|
198 |
+
- 1.0
|
199 |
+
- 1.0
|
200 |
+
- 1.0
|
201 |
+
FOCAL_LOSS_ALPHA: 0.25
|
202 |
+
FOCAL_LOSS_GAMMA: 2.0
|
203 |
+
IN_FEATURES:
|
204 |
+
- p3
|
205 |
+
- p4
|
206 |
+
- p5
|
207 |
+
- p6
|
208 |
+
- p7
|
209 |
+
IOU_LABELS:
|
210 |
+
- 0
|
211 |
+
- -1
|
212 |
+
- 1
|
213 |
+
IOU_THRESHOLDS:
|
214 |
+
- 0.4
|
215 |
+
- 0.5
|
216 |
+
NMS_THRESH_TEST: 0.5
|
217 |
+
NORM: ''
|
218 |
+
NUM_CLASSES: 80
|
219 |
+
NUM_CONVS: 4
|
220 |
+
PRIOR_PROB: 0.01
|
221 |
+
SCORE_THRESH_TEST: 0.05
|
222 |
+
SMOOTH_L1_LOSS_BETA: 0.1
|
223 |
+
TOPK_CANDIDATES_TEST: 1000
|
224 |
+
ROI_BOX_CASCADE_HEAD:
|
225 |
+
BBOX_REG_WEIGHTS:
|
226 |
+
- &id001
|
227 |
+
- 10.0
|
228 |
+
- 10.0
|
229 |
+
- 5.0
|
230 |
+
- 5.0
|
231 |
+
- - 20.0
|
232 |
+
- 20.0
|
233 |
+
- 10.0
|
234 |
+
- 10.0
|
235 |
+
- - 30.0
|
236 |
+
- 30.0
|
237 |
+
- 15.0
|
238 |
+
- 15.0
|
239 |
+
IOUS:
|
240 |
+
- 0.5
|
241 |
+
- 0.6
|
242 |
+
- 0.7
|
243 |
+
ROI_BOX_HEAD:
|
244 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
245 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
246 |
+
BBOX_REG_WEIGHTS: *id001
|
247 |
+
CLS_AGNOSTIC_BBOX_REG: false
|
248 |
+
CONV_DIM: 256
|
249 |
+
FC_DIM: 1024
|
250 |
+
FED_LOSS_FREQ_WEIGHT_POWER: 0.5
|
251 |
+
FED_LOSS_NUM_CLASSES: 50
|
252 |
+
NAME: ''
|
253 |
+
NORM: ''
|
254 |
+
NUM_CONV: 0
|
255 |
+
NUM_FC: 0
|
256 |
+
POOLER_RESOLUTION: 14
|
257 |
+
POOLER_SAMPLING_RATIO: 0
|
258 |
+
POOLER_TYPE: ROIAlignV2
|
259 |
+
SMOOTH_L1_BETA: 0.0
|
260 |
+
TRAIN_ON_PRED_BOXES: false
|
261 |
+
USE_FED_LOSS: false
|
262 |
+
USE_SIGMOID_CE: false
|
263 |
+
ROI_HEADS:
|
264 |
+
BATCH_SIZE_PER_IMAGE: 512
|
265 |
+
IN_FEATURES:
|
266 |
+
- res4
|
267 |
+
IOU_LABELS:
|
268 |
+
- 0
|
269 |
+
- 1
|
270 |
+
IOU_THRESHOLDS:
|
271 |
+
- 0.5
|
272 |
+
NAME: Res5ROIHeads
|
273 |
+
NMS_THRESH_TEST: 0.5
|
274 |
+
NUM_CLASSES: 80
|
275 |
+
POSITIVE_FRACTION: 0.25
|
276 |
+
PROPOSAL_APPEND_GT: true
|
277 |
+
SCORE_THRESH_TEST: 0.05
|
278 |
+
ROI_KEYPOINT_HEAD:
|
279 |
+
CONV_DIMS:
|
280 |
+
- 512
|
281 |
+
- 512
|
282 |
+
- 512
|
283 |
+
- 512
|
284 |
+
- 512
|
285 |
+
- 512
|
286 |
+
- 512
|
287 |
+
- 512
|
288 |
+
LOSS_WEIGHT: 1.0
|
289 |
+
MIN_KEYPOINTS_PER_IMAGE: 1
|
290 |
+
NAME: KRCNNConvDeconvUpsampleHead
|
291 |
+
NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: true
|
292 |
+
NUM_KEYPOINTS: 17
|
293 |
+
POOLER_RESOLUTION: 14
|
294 |
+
POOLER_SAMPLING_RATIO: 0
|
295 |
+
POOLER_TYPE: ROIAlignV2
|
296 |
+
ROI_MASK_HEAD:
|
297 |
+
CLS_AGNOSTIC_MASK: false
|
298 |
+
CONV_DIM: 256
|
299 |
+
NAME: MaskRCNNConvUpsampleHead
|
300 |
+
NORM: ''
|
301 |
+
NUM_CONV: 0
|
302 |
+
POOLER_RESOLUTION: 14
|
303 |
+
POOLER_SAMPLING_RATIO: 0
|
304 |
+
POOLER_TYPE: ROIAlignV2
|
305 |
+
RPN:
|
306 |
+
BATCH_SIZE_PER_IMAGE: 256
|
307 |
+
BBOX_REG_LOSS_TYPE: smooth_l1
|
308 |
+
BBOX_REG_LOSS_WEIGHT: 1.0
|
309 |
+
BBOX_REG_WEIGHTS: *id002
|
310 |
+
BOUNDARY_THRESH: -1
|
311 |
+
CONV_DIMS:
|
312 |
+
- -1
|
313 |
+
HEAD_NAME: StandardRPNHead
|
314 |
+
IN_FEATURES:
|
315 |
+
- res4
|
316 |
+
IOU_LABELS:
|
317 |
+
- 0
|
318 |
+
- -1
|
319 |
+
- 1
|
320 |
+
IOU_THRESHOLDS:
|
321 |
+
- 0.3
|
322 |
+
- 0.7
|
323 |
+
LOSS_WEIGHT: 1.0
|
324 |
+
NMS_THRESH: 0.7
|
325 |
+
POSITIVE_FRACTION: 0.5
|
326 |
+
POST_NMS_TOPK_TEST: 1000
|
327 |
+
POST_NMS_TOPK_TRAIN: 2000
|
328 |
+
PRE_NMS_TOPK_TEST: 6000
|
329 |
+
PRE_NMS_TOPK_TRAIN: 12000
|
330 |
+
SMOOTH_L1_BETA: 0.0
|
331 |
+
SEM_SEG_HEAD:
|
332 |
+
ASPP_CHANNELS: 256
|
333 |
+
ASPP_DILATIONS:
|
334 |
+
- 6
|
335 |
+
- 12
|
336 |
+
- 18
|
337 |
+
ASPP_DROPOUT: 0.1
|
338 |
+
COMMON_STRIDE: 4
|
339 |
+
CONVS_DIM: 256
|
340 |
+
DEFORMABLE_TRANSFORMER_ENCODER_IN_FEATURES:
|
341 |
+
- res3
|
342 |
+
- res4
|
343 |
+
- res5
|
344 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_HEADS: 8
|
345 |
+
DEFORMABLE_TRANSFORMER_ENCODER_N_POINTS: 4
|
346 |
+
IGNORE_VALUE: 255
|
347 |
+
IN_FEATURES:
|
348 |
+
- res2
|
349 |
+
- res3
|
350 |
+
- res4
|
351 |
+
- res5
|
352 |
+
LOSS_TYPE: hard_pixel_mining
|
353 |
+
LOSS_WEIGHT: 1.0
|
354 |
+
MASK_DIM: 256
|
355 |
+
NAME: MaskFormerHead
|
356 |
+
NORM: GN
|
357 |
+
NUM_CLASSES: 150
|
358 |
+
PIXEL_DECODER_NAME: MSDeformAttnPixelDecoder
|
359 |
+
PROJECT_CHANNELS:
|
360 |
+
- 48
|
361 |
+
PROJECT_FEATURES:
|
362 |
+
- res2
|
363 |
+
TRANSFORMER_ENC_LAYERS: 6
|
364 |
+
USE_DEPTHWISE_SEPARABLE_CONV: false
|
365 |
+
SWIN:
|
366 |
+
APE: false
|
367 |
+
ATTN_DROP_RATE: 0.0
|
368 |
+
DEPTHS:
|
369 |
+
- 2
|
370 |
+
- 2
|
371 |
+
- 6
|
372 |
+
- 2
|
373 |
+
DROP_PATH_RATE: 0.3
|
374 |
+
DROP_RATE: 0.0
|
375 |
+
EMBED_DIM: 96
|
376 |
+
MLP_RATIO: 4.0
|
377 |
+
NUM_HEADS:
|
378 |
+
- 3
|
379 |
+
- 6
|
380 |
+
- 12
|
381 |
+
- 24
|
382 |
+
OUT_FEATURES:
|
383 |
+
- res2
|
384 |
+
- res3
|
385 |
+
- res4
|
386 |
+
- res5
|
387 |
+
PATCH_NORM: true
|
388 |
+
PATCH_SIZE: 4
|
389 |
+
PRETRAIN_IMG_SIZE: 224
|
390 |
+
QKV_BIAS: true
|
391 |
+
QK_SCALE: null
|
392 |
+
USE_CHECKPOINT: false
|
393 |
+
WINDOW_SIZE: 7
|
394 |
+
WEIGHTS: ./output/newss/100-5_1_1+2_vq3/step10/model_final.pth
|
395 |
+
OUTPUT_DIR: ./output/newss/100-5_1_1+2_vq3/step11
|
396 |
+
SEED: -1
|
397 |
+
SOLVER:
|
398 |
+
AMP:
|
399 |
+
ENABLED: true
|
400 |
+
BACKBONE_MULTIPLIER: 0.1
|
401 |
+
BASE_LR: 5.0e-05
|
402 |
+
BASE_LR_END: 0.0
|
403 |
+
BIAS_LR_FACTOR: 1.0
|
404 |
+
CHECKPOINT_PERIOD: 10000
|
405 |
+
CLIP_GRADIENTS:
|
406 |
+
CLIP_TYPE: full_model
|
407 |
+
CLIP_VALUE: 0.01
|
408 |
+
ENABLED: true
|
409 |
+
NORM_TYPE: 2.0
|
410 |
+
GAMMA: 0.1
|
411 |
+
IMS_PER_BATCH: 8
|
412 |
+
LR_SCHEDULER_NAME: WarmupMultiStepLR
|
413 |
+
MAX_ITER: 5000
|
414 |
+
MOMENTUM: 0.9
|
415 |
+
NESTEROV: false
|
416 |
+
NUM_DECAYS: 3
|
417 |
+
OPTIMIZER: ADAMW
|
418 |
+
POLY_LR_CONSTANT_ENDING: 0.0
|
419 |
+
POLY_LR_POWER: 0.9
|
420 |
+
REFERENCE_WORLD_SIZE: 0
|
421 |
+
RESCALE_INTERVAL: false
|
422 |
+
STEPS:
|
423 |
+
- 4250
|
424 |
+
- 4750
|
425 |
+
WARMUP_FACTOR: 1.0
|
426 |
+
WARMUP_ITERS: 0
|
427 |
+
WARMUP_METHOD: linear
|
428 |
+
WEIGHT_DECAY: 0.05
|
429 |
+
WEIGHT_DECAY_BIAS: null
|
430 |
+
WEIGHT_DECAY_EMBED: 0.0
|
431 |
+
WEIGHT_DECAY_NORM: 0.0
|
432 |
+
TEST:
|
433 |
+
AUG:
|
434 |
+
ENABLED: false
|
435 |
+
FLIP: true
|
436 |
+
MAX_SIZE: 3584
|
437 |
+
MIN_SIZES:
|
438 |
+
- 256
|
439 |
+
- 384
|
440 |
+
- 512
|
441 |
+
- 640
|
442 |
+
- 768
|
443 |
+
- 896
|
444 |
+
DETECTIONS_PER_IMAGE: 100
|
445 |
+
EVAL_PERIOD: 5000
|
446 |
+
EXPECTED_RESULTS: []
|
447 |
+
KEYPOINT_OKS_SIGMAS: []
|
448 |
+
PRECISE_BN:
|
449 |
+
ENABLED: false
|
450 |
+
NUM_ITER: 200
|
451 |
+
VERSION: 2
|
452 |
+
VIS_PERIOD: 0
|
ss_100-5_38/step11/fake_query.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f1063758c31c11654d047d1c56383d18070f43e210559b763b42eb04e7c28b6e
|
3 |
+
size 27746927
|
ss_100-5_38/step11/last_checkpoint
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
model_final.pth
|
ss_100-5_38/step11/metrics.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
ss_100-5_38/step11/model_final.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d6e378e5efe890d3c72dd23504454a53c62ffa5552115346e62a279f74fc6ab1
|
3 |
+
size 762308965
|
ss_100-5_38/step11/psd_distribution.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
[1495.0, 418.0, 363.0, 1229.0, 284.0, 701.0, 185.0, 328.0, 679.0, 126.0, 507.0, 258.0, 441.0, 20.0, 353.0, 861.0, 31.0, 493.0, 398.0, 658.0, 238.0, 8.0, 378.0, 204.0, 243.0, 0.0, 8.0, 206.0, 163.0, 1.0, 90.0, 13.0, 48.0, 34.0, 15.0, 21.0, 541.0, 61.0, 38.0, 244.0, 6.0, 460.0, 77.0, 162.0, 65.0, 30.0, 2.0, 222.0, 8.0, 77.0, 61.0, 7.0, 25.0, 48.0, 2.0, 9.0, 28.0, 212.0, 28.0, 16.0, 0.0, 14.0, 31.0, 34.0, 85.0, 85.0, 268.0, 347.0, 1.0, 19.0, 173.0, 94.0, 9.0, 58.0, 19.0, 14.0, 2.0, 3.0, 5.0, 0.0, 13.0, 96.0, 217.0, 23.0, 26.0, 90.0, 53.0, 103.0, 5.0, 69.0, 2.0, 0.0, 18.0, 102.0, 0.0, 30.0, 6.0, 41.0, 424.0, 13.0, 297.0, 1.0, 182.0, 5.0, 7.0, 24.0, 9.0, 30.0, 800.0, 18.0, 567.0, 0.0, 540.0, 0.0, 6.0, 186.0, 50.0, 8.0, 55.0, 24.0, 15.0, 282.0, 1.0, 59.0, 103.0, 193.0, 0.0, 44.0, 4.0, 39.0, 5.0, 39.0, 19.0, 35.0, 234.0, 234.0, 18.0, 17.0, 104.0, 42.0, 8.0, 49.0, 207.0, 19.0, 40.0, 0.0, 0.0, 0.0, 0.0, 0.0]
|