Spaces:
Running
on
Zero
Running
on
Zero
add gradio ui
Browse files
README.md
CHANGED
@@ -11,4 +11,21 @@ license: apache-2.0
|
|
11 |
short_description: RTMO PyTorch Checkpoint Tester
|
12 |
---
|
13 |
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
short_description: RTMO PyTorch Checkpoint Tester
|
12 |
---
|
13 |
|
14 |
+
# RTMO PyTorch Checkpoint Tester
|
15 |
+
|
16 |
+
This HuggingFace Space runs the RTMO (Real-Time Multi-Person) 2D pose estimation model from OpenMMLab.
|
17 |
+
|
18 |
+
## Usage
|
19 |
+
1. Upload an image via the Gradio UI.
|
20 |
+
2. (Optional) Provide a path or URL to your own RTMO PyTorch `.pth` checkpoint. If left blank, the default pretrained weights will be used.
|
21 |
+
3. Click **Submit**. The annotated image with keypoints will be displayed.
|
22 |
+
|
23 |
+
## Files
|
24 |
+
- **app.py**: Gradio application script that loads the RTMO model, runs inference, and displays results.
|
25 |
+
- **requirements.txt**: Python dependencies, including the patched MMCV build and MMPose.
|
26 |
+
|
27 |
+
## Model
|
28 |
+
We use the `rtmo` alias defined in MMPoseβs model zoo. To override, upload your own checkpoint.
|
29 |
+
|
30 |
+
## Development
|
31 |
+
If you need to update dependencies or change the model, modify `requirements.txt` and `app.py` accordingly.
|
app.py
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
import spaces
|
3 |
+
import os, sys, importlib.util, re
|
4 |
+
import gradio as gr
|
5 |
+
from PIL import Image
|
6 |
+
|
7 |
+
# βββ Monkey-patch mmdet to remove its mmcv-version assertion βββ
|
8 |
+
spec = importlib.util.find_spec('mmdet')
|
9 |
+
if spec and spec.origin:
|
10 |
+
src = open(spec.origin, encoding='utf-8').read()
|
11 |
+
# strip out the mmcv_minimum_versionβ¦assertβ¦ block up to __all__
|
12 |
+
patched = re.sub(r'(?ms)^[ \t]*mmcv_minimum_version.*?^__all__', '__all__', src)
|
13 |
+
m = importlib.util.module_from_spec(spec)
|
14 |
+
m.__loader__ = spec.loader
|
15 |
+
m.__file__ = spec.origin
|
16 |
+
m.__path__ = spec.submodule_search_locations
|
17 |
+
sys.modules['mmdet'] = m
|
18 |
+
exec(compile(patched, spec.origin, 'exec'), m.__dict__)
|
19 |
+
|
20 |
+
from mmpose.apis.inferencers import MMPoseInferencer
|
21 |
+
|
22 |
+
# βββ Initialize inferencer with default RTMO 2D model βββ
|
23 |
+
def load_inferencer(checkpoint_path=None, device=None):
|
24 |
+
kwargs = {'pose2d': 'rtmo', 'scope': 'mmpose', 'device': device, 'det_cat_ids': [0]}
|
25 |
+
if checkpoint_path:
|
26 |
+
kwargs['pose2d_weights'] = checkpoint_path
|
27 |
+
return MMPoseInferencer(**kwargs)
|
28 |
+
|
29 |
+
# βββ Gradio prediction function βββ
|
30 |
+
@spaces.GPU()
|
31 |
+
def predict(image: Image.Image, checkpoint: str = None):
|
32 |
+
# save upload to temp file
|
33 |
+
inp_path = "/tmp/upload.jpg"
|
34 |
+
image.save(inp_path)
|
35 |
+
|
36 |
+
vis_dir = "/tmp/vis"
|
37 |
+
os.makedirs(vis_dir, exist_ok=True)
|
38 |
+
|
39 |
+
inferencer = load_inferencer(checkpoint_path=checkpoint, device=None)
|
40 |
+
# run inference & visualization
|
41 |
+
for result in inferencer(
|
42 |
+
inputs=inp_path,
|
43 |
+
bbox_thr=0.1,
|
44 |
+
nms_thr=0.65,
|
45 |
+
pose_based_nms=True,
|
46 |
+
show=False,
|
47 |
+
vis_out_dir=vis_dir,
|
48 |
+
):
|
49 |
+
pass
|
50 |
+
|
51 |
+
# return the first visualization
|
52 |
+
out_files = sorted(os.listdir(vis_dir))
|
53 |
+
if out_files:
|
54 |
+
return Image.open(os.path.join(vis_dir, out_files[0]))
|
55 |
+
return None
|
56 |
+
|
57 |
+
# βββ Build Gradio Interface βββ
|
58 |
+
demo = gr.Interface(
|
59 |
+
fn=predict,
|
60 |
+
inputs=[
|
61 |
+
gr.inputs.Image(type="pil", label="Upload Image"),
|
62 |
+
gr.inputs.Text(label="RTMO PyTorch Checkpoint Path (optional)")
|
63 |
+
],
|
64 |
+
outputs=gr.outputs.Image(type="pil", label="Annotated Image"),
|
65 |
+
title="RTMO Pose Demo",
|
66 |
+
description="Upload an image, optionally supply a RTMO .pth checkpoint, and see 2D pose annotation.",
|
67 |
+
)
|
68 |
+
|
69 |
+
def main():
|
70 |
+
demo.launch()
|
71 |
+
|
72 |
+
if __name__ == "__main__":
|
73 |
+
main()
|