🔧 [Add] v9-c-seg config file
Browse files- yolo/config/model/v9-c-seg.yaml +151 -0
yolo/config/model/v9-c-seg.yaml
ADDED
@@ -0,0 +1,151 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
name: v9-c-seg
|
2 |
+
|
3 |
+
anchor:
|
4 |
+
reg_max: 16
|
5 |
+
strides: [8, 16, 32]
|
6 |
+
|
7 |
+
model:
|
8 |
+
backbone:
|
9 |
+
- Conv:
|
10 |
+
args: {out_channels: 64, kernel_size: 3, stride: 2}
|
11 |
+
source: 0
|
12 |
+
- Conv:
|
13 |
+
args: {out_channels: 128, kernel_size: 3, stride: 2}
|
14 |
+
- RepNCSPELAN:
|
15 |
+
args: {out_channels: 256, part_channels: 128}
|
16 |
+
|
17 |
+
- ADown:
|
18 |
+
args: {out_channels: 256}
|
19 |
+
- RepNCSPELAN:
|
20 |
+
args: {out_channels: 512, part_channels: 256}
|
21 |
+
tags: B3
|
22 |
+
|
23 |
+
- ADown:
|
24 |
+
args: {out_channels: 512}
|
25 |
+
- RepNCSPELAN:
|
26 |
+
args: {out_channels: 512, part_channels: 512}
|
27 |
+
tags: B4
|
28 |
+
|
29 |
+
- ADown:
|
30 |
+
args: {out_channels: 512}
|
31 |
+
- RepNCSPELAN:
|
32 |
+
args: {out_channels: 512, part_channels: 512}
|
33 |
+
tags: B5
|
34 |
+
|
35 |
+
neck:
|
36 |
+
- SPPELAN:
|
37 |
+
args: {out_channels: 512}
|
38 |
+
tags: N3
|
39 |
+
|
40 |
+
- UpSample:
|
41 |
+
args: {scale_factor: 2, mode: nearest}
|
42 |
+
- Concat:
|
43 |
+
source: [-1, B4]
|
44 |
+
- RepNCSPELAN:
|
45 |
+
args: {out_channels: 512, part_channels: 512}
|
46 |
+
tags: N4
|
47 |
+
|
48 |
+
- UpSample:
|
49 |
+
args: {scale_factor: 2, mode: nearest}
|
50 |
+
- Concat:
|
51 |
+
source: [-1, B3]
|
52 |
+
|
53 |
+
head:
|
54 |
+
- RepNCSPELAN:
|
55 |
+
args: {out_channels: 256, part_channels: 256}
|
56 |
+
tags: P3
|
57 |
+
|
58 |
+
- ADown:
|
59 |
+
args: {out_channels: 256}
|
60 |
+
- Concat:
|
61 |
+
source: [-1, N4]
|
62 |
+
- RepNCSPELAN:
|
63 |
+
args: {out_channels: 512, part_channels: 512}
|
64 |
+
tags: P4
|
65 |
+
|
66 |
+
- ADown:
|
67 |
+
args: {out_channels: 512}
|
68 |
+
- Concat:
|
69 |
+
source: [-1, N3]
|
70 |
+
- RepNCSPELAN:
|
71 |
+
args: {out_channels: 512, part_channels: 512}
|
72 |
+
tags: P5
|
73 |
+
|
74 |
+
detection:
|
75 |
+
- RepNCSPELAN:
|
76 |
+
source: P3
|
77 |
+
args: {out_channels: 256, part_channels: 256, csp_args: {repeat_num: 2}}
|
78 |
+
- UpSample:
|
79 |
+
args: {scale_factor: 2, mode: nearest}
|
80 |
+
- Conv:
|
81 |
+
args: {out_channels: 256, kernel_size: 3}
|
82 |
+
|
83 |
+
- MultiheadSegmentation:
|
84 |
+
source: [P3, P4, P5, -1]
|
85 |
+
args: {num_maskes: 32}
|
86 |
+
tags: Main
|
87 |
+
output: True
|
88 |
+
|
89 |
+
auxiliary:
|
90 |
+
- CBLinear:
|
91 |
+
source: B3
|
92 |
+
args: {out_channels: [256]}
|
93 |
+
tags: R3
|
94 |
+
- CBLinear:
|
95 |
+
source: B4
|
96 |
+
args: {out_channels: [256, 512]}
|
97 |
+
tags: R4
|
98 |
+
- CBLinear:
|
99 |
+
source: B5
|
100 |
+
args: {out_channels: [256, 512, 512]}
|
101 |
+
tags: R5
|
102 |
+
|
103 |
+
- Conv:
|
104 |
+
args: {out_channels: 64, kernel_size: 3, stride: 2}
|
105 |
+
source: 0
|
106 |
+
- Conv:
|
107 |
+
args: {out_channels: 128, kernel_size: 3, stride: 2}
|
108 |
+
- RepNCSPELAN:
|
109 |
+
args: {out_channels: 256, part_channels: 128}
|
110 |
+
|
111 |
+
- ADown:
|
112 |
+
args: {out_channels: 256}
|
113 |
+
- CBFuse:
|
114 |
+
source: [R3, R4, R5, -1]
|
115 |
+
args: {index: [0, 0, 0]}
|
116 |
+
- RepNCSPELAN:
|
117 |
+
args: {out_channels: 512, part_channels: 256}
|
118 |
+
tags: A3
|
119 |
+
|
120 |
+
- ADown:
|
121 |
+
args: {out_channels: 512}
|
122 |
+
- CBFuse:
|
123 |
+
source: [R4, R5, -1]
|
124 |
+
args: {index: [1, 1]}
|
125 |
+
- RepNCSPELAN:
|
126 |
+
args: {out_channels: 512, part_channels: 512}
|
127 |
+
tags: A4
|
128 |
+
|
129 |
+
- ADown:
|
130 |
+
args: {out_channels: 512}
|
131 |
+
- CBFuse:
|
132 |
+
source: [R5, -1]
|
133 |
+
args: {index: [2]}
|
134 |
+
- RepNCSPELAN:
|
135 |
+
args: {out_channels: 512, part_channels: 512}
|
136 |
+
tags: A5
|
137 |
+
|
138 |
+
- RepNCSPELAN:
|
139 |
+
source: A3
|
140 |
+
args: {out_channels: 512, part_channels: 256, csp_args: {repeat_num: 2}}
|
141 |
+
|
142 |
+
- UpSample:
|
143 |
+
args: {scale_factor: 2, mode: nearest}
|
144 |
+
- Conv:
|
145 |
+
args: {out_channels: 256, kernel_size: 3}
|
146 |
+
|
147 |
+
- MultiheadSegmentation:
|
148 |
+
source: [A3, A4, A5, -1]
|
149 |
+
args: {num_maskes: 32}
|
150 |
+
tags: AUX
|
151 |
+
output: True
|