Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Commit
·
b38382a
1
Parent(s):
d753a84
app.py added
Browse files- README.md +1 -0
- app.py +200 -0
- requirements.txt +1 -0
README.md
CHANGED
@@ -8,6 +8,7 @@ sdk_version: 4.15.0
|
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
license: mit
|
|
|
11 |
---
|
12 |
|
13 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
license: mit
|
11 |
+
python_version: 3.8
|
12 |
---
|
13 |
|
14 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
app.py
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright (c) Microsoft Corporation. All rights reserved.
|
2 |
+
# Licensed under the MIT License.
|
3 |
+
|
4 |
+
""" Demo using Gradio interface"""
|
5 |
+
|
6 |
+
#%%
|
7 |
+
# Importing basic libraries
|
8 |
+
import os
|
9 |
+
import time
|
10 |
+
from PIL import Image
|
11 |
+
import supervision as sv
|
12 |
+
import gradio as gr
|
13 |
+
from zipfile import ZipFile
|
14 |
+
from torch.utils.data import DataLoader
|
15 |
+
|
16 |
+
#%%
|
17 |
+
# Importing the models, dataset, transformations, and utility functions from PytorchWildlife
|
18 |
+
from PytorchWildlife.models import detection as pw_detection
|
19 |
+
from PytorchWildlife.models import classification as pw_classification
|
20 |
+
from PytorchWildlife.data import transforms as pw_trans
|
21 |
+
from PytorchWildlife.data import datasets as pw_data
|
22 |
+
from PytorchWildlife import utils as pw_utils
|
23 |
+
|
24 |
+
#%%
|
25 |
+
# Setting the device to use for computations ('cuda' indicates GPU)
|
26 |
+
DEVICE = "cuda"
|
27 |
+
# Initializing a supervision box annotator for visualizing detections
|
28 |
+
box_annotator = sv.BoxAnnotator(thickness=4, text_thickness=4, text_scale=2)
|
29 |
+
|
30 |
+
# Initializing the detection and classification models
|
31 |
+
detection_model = None
|
32 |
+
classification_model = None
|
33 |
+
|
34 |
+
# Defining transformations for detection and classification
|
35 |
+
trans_det = None
|
36 |
+
trans_clf = None
|
37 |
+
|
38 |
+
#%% Defining functions for different detection scenarios
|
39 |
+
def load_models(det, clf):
|
40 |
+
|
41 |
+
global detection_model, classification_model, trans_det, trans_clf
|
42 |
+
|
43 |
+
detection_model = pw_detection.__dict__[det](device=DEVICE, pretrained=True)
|
44 |
+
if clf != "None":
|
45 |
+
classification_model = pw_classification.__dict__[clf](device=DEVICE, pretrained=True)
|
46 |
+
|
47 |
+
trans_det = pw_trans.MegaDetector_v5_Transform(target_size=detection_model.IMAGE_SIZE,
|
48 |
+
stride=detection_model.STRIDE)
|
49 |
+
trans_clf = pw_trans.Classification_Inference_Transform(target_size=224)
|
50 |
+
|
51 |
+
return "Loaded Detector: {}. Loaded Classifier: {}".format(det, clf)
|
52 |
+
|
53 |
+
|
54 |
+
def single_image_detection(input_img, det_conf_thres, clf_conf_thres, img_index=None):
|
55 |
+
"""Performs detection on a single image and returns an annotated image.
|
56 |
+
|
57 |
+
Args:
|
58 |
+
input_img (np.ndarray): Input image in numpy array format defaulted by Gradio.
|
59 |
+
det_conf_thre (float): Confidence threshold for detection.
|
60 |
+
clf_conf_thre (float): Confidence threshold for classification.
|
61 |
+
img_index: Image index identifier.
|
62 |
+
Returns:
|
63 |
+
annotated_img (PIL.Image.Image): Annotated image with bounding box instances.
|
64 |
+
"""
|
65 |
+
results_det = detection_model.single_image_detection(trans_det(input_img),
|
66 |
+
input_img.shape,
|
67 |
+
img_path=img_index,
|
68 |
+
conf_thres=det_conf_thres)
|
69 |
+
if classification_model is not None:
|
70 |
+
labels = []
|
71 |
+
for xyxy, det_id in zip(results_det["detections"].xyxy, results_det["detections"].class_id):
|
72 |
+
# Only run classifier when detection class is animal
|
73 |
+
if det_id == 0:
|
74 |
+
cropped_image = sv.crop_image(image=input_img, xyxy=xyxy)
|
75 |
+
results_clf = classification_model.single_image_classification(trans_clf(Image.fromarray(cropped_image)))
|
76 |
+
labels.append("{} {:.2f}".format(results_clf["prediction"] if results_clf["confidence"] > clf_conf_thres else "Unknown",
|
77 |
+
results_clf["confidence"]))
|
78 |
+
else:
|
79 |
+
labels = results_det["labels"]
|
80 |
+
else:
|
81 |
+
labels = results_det["labels"]
|
82 |
+
annotated_img = box_annotator.annotate(scene=input_img, detections=results_det["detections"], labels=labels)
|
83 |
+
return annotated_img
|
84 |
+
|
85 |
+
|
86 |
+
def batch_detection(zip_file, det_conf_thres):
|
87 |
+
"""Perform detection on a batch of images from a zip file and return path to results JSON.
|
88 |
+
|
89 |
+
Args:
|
90 |
+
zip_file (File): Zip file containing images.
|
91 |
+
det_conf_thre (float): Confidence threshold for detection.
|
92 |
+
clf_conf_thre (float): Confidence threshold for classification.
|
93 |
+
|
94 |
+
Returns:
|
95 |
+
json_save_path (str): Path to the JSON file containing detection results.
|
96 |
+
"""
|
97 |
+
extract_path = os.path.join("..","temp","zip_upload")
|
98 |
+
json_save_path = os.path.join(extract_path, "results.json")
|
99 |
+
with ZipFile(zip_file.name) as zfile:
|
100 |
+
zfile.extractall(extract_path)
|
101 |
+
#tgt_folder_path = os.path.join(extract_path, zip_file.name.rsplit(os.sep, 1)[1].rstrip(".zip"))
|
102 |
+
tgt_folder_path = os.path.join(extract_path)
|
103 |
+
det_dataset = pw_data.DetectionImageFolder(tgt_folder_path, transform=trans_det)
|
104 |
+
det_loader = DataLoader(det_dataset, batch_size=32, shuffle=False,
|
105 |
+
pin_memory=True, num_workers=8, drop_last=False)
|
106 |
+
det_results = detection_model.batch_image_detection(det_loader, conf_thres=det_conf_thres, id_strip=tgt_folder_path)
|
107 |
+
|
108 |
+
if classification_model is not None:
|
109 |
+
clf_dataset = pw_data.DetectionCrops(
|
110 |
+
det_results,
|
111 |
+
transform=pw_trans.Classification_Inference_Transform(target_size=224),
|
112 |
+
path_head=tgt_folder_path
|
113 |
+
)
|
114 |
+
clf_loader = DataLoader(clf_dataset, batch_size=32, shuffle=False,
|
115 |
+
pin_memory=True, num_workers=8, drop_last=False)
|
116 |
+
clf_results = classification_model.batch_image_classification(clf_loader, id_strip=tgt_folder_path)
|
117 |
+
pw_utils.save_detection_classification_json(det_results=det_results,
|
118 |
+
clf_results=clf_results,
|
119 |
+
det_categories=detection_model.CLASS_NAMES,
|
120 |
+
clf_categories=classification_model.CLASS_NAMES,
|
121 |
+
output_path=json_save_path)
|
122 |
+
else:
|
123 |
+
pw_utils.save_detection_json(det_results, json_save_path, categories=detection_model.CLASS_NAMES)
|
124 |
+
|
125 |
+
return json_save_path
|
126 |
+
|
127 |
+
|
128 |
+
def video_detection(video, det_conf_thres, clf_conf_thres, target_fps):
|
129 |
+
"""Perform detection on a video and return path to processed video.
|
130 |
+
|
131 |
+
Args:
|
132 |
+
video (str): Video source path.
|
133 |
+
det_conf_thre (float): Confidence threshold for detection.
|
134 |
+
clf_conf_thre (float): Confidence threshold for classification.
|
135 |
+
|
136 |
+
"""
|
137 |
+
def callback(frame, index):
|
138 |
+
annotated_frame = single_image_detection(frame,
|
139 |
+
img_index=index,
|
140 |
+
det_conf_thres=det_conf_thres,
|
141 |
+
clf_conf_thres=clf_conf_thres)
|
142 |
+
return annotated_frame
|
143 |
+
|
144 |
+
target_path = "../temp/video_detection.mp4"
|
145 |
+
pw_utils.process_video(source_path=video, target_path=target_path,
|
146 |
+
callback=callback, target_fps=target_fps)
|
147 |
+
return target_path
|
148 |
+
|
149 |
+
#%% Building Gradio UI
|
150 |
+
|
151 |
+
with gr.Blocks() as demo:
|
152 |
+
gr.Markdown("# Pytorch-Wildlife Demo.")
|
153 |
+
with gr.Row():
|
154 |
+
det_drop = gr.Dropdown(
|
155 |
+
["MegaDetectorV5"],
|
156 |
+
label="Detection model",
|
157 |
+
info="Will add more detection models!",
|
158 |
+
value="MegaDetectorV5"
|
159 |
+
)
|
160 |
+
clf_drop = gr.Dropdown(
|
161 |
+
["None", "AI4GOpossum", "AI4GAmazonRainforest"],
|
162 |
+
label="Classification model",
|
163 |
+
info="Will add more classification models!",
|
164 |
+
value="None"
|
165 |
+
)
|
166 |
+
with gr.Column():
|
167 |
+
load_but = gr.Button("Load Models!")
|
168 |
+
load_out = gr.Text("NO MODEL LOADED!!", label="Loaded models:")
|
169 |
+
with gr.Tab("Single Image Process"):
|
170 |
+
with gr.Row():
|
171 |
+
with gr.Column():
|
172 |
+
sgl_in = gr.Image()
|
173 |
+
sgl_conf_sl_det = gr.Slider(0, 1, label="Detection Confidence Threshold", value=0.2)
|
174 |
+
sgl_conf_sl_clf = gr.Slider(0, 1, label="Classification Confidence Threshold", value=0.7)
|
175 |
+
sgl_out = gr.Image()
|
176 |
+
sgl_but = gr.Button("Detect Animals!")
|
177 |
+
with gr.Tab("Batch Image Process"):
|
178 |
+
with gr.Row():
|
179 |
+
with gr.Column():
|
180 |
+
bth_in = gr.File(label="Upload zip file.")
|
181 |
+
bth_conf_sl = gr.Slider(0, 1, label="Detection Confidence Threshold", value=0.2)
|
182 |
+
bth_out = gr.File(label="Detection Results JSON.", height=200)
|
183 |
+
bth_but = gr.Button("Detect Animals!")
|
184 |
+
with gr.Tab("Single Video Process"):
|
185 |
+
with gr.Row():
|
186 |
+
with gr.Column():
|
187 |
+
vid_in = gr.Video(label="Upload a video.")
|
188 |
+
vid_conf_sl_det = gr.Slider(0, 1, label="Detection Confidence Threshold", value=0.2)
|
189 |
+
vid_conf_sl_clf = gr.Slider(0, 1, label="Classification Confidence Threshold", value=0.7)
|
190 |
+
vid_fr = gr.Dropdown([5, 10, 30], label="Output video framerate", value=30)
|
191 |
+
vid_out = gr.Video()
|
192 |
+
vid_but = gr.Button("Detect Animals!")
|
193 |
+
|
194 |
+
load_but.click(load_models, inputs=[det_drop, clf_drop], outputs=load_out)
|
195 |
+
sgl_but.click(single_image_detection, inputs=[sgl_in, sgl_conf_sl_det, sgl_conf_sl_clf], outputs=sgl_out)
|
196 |
+
bth_but.click(batch_detection, inputs=[bth_in, bth_conf_sl], outputs=bth_out)
|
197 |
+
vid_but.click(video_detection, inputs=[vid_in, vid_conf_sl_det, vid_conf_sl_clf, vid_fr], outputs=vid_out)
|
198 |
+
|
199 |
+
if __name__ == "__main__":
|
200 |
+
demo.launch(share=True)
|
requirements.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
PytorchWildlife
|