cv-zmh / human-action-recognition Goto Github PK

Hi,

First of all, thank you so much for sharing this wonderful work. Really appreciate it.

I have a question regarding the demo.py. Could you help me?

In the demo.py, we initialize the pose_estimator, tracker, and action_classifier before we loop over the video frame.

But, if I try to initialize those 3 above functions inside the loop.

I couldn't get a proper prediction.

I only can get the correct prediction if I initialize them before the loop.

May I ask why I can't get a proper prediction?

Thank you so much for reading this message.

Looking forward to hearing from you.

While running the demo.py on wideresnet_mars.pth or siamesenet_mars.pth the terminal gets stuck unless i close the video window, then % increases and then after some windows are closed it starts giving the below error.
Any help is much appreciated, Thanks!!

(humanaction) C:\Users\Utkarsh.Singh\Documents\Utkarsh\PoseTracking\human-action-recognition\src>python demo.py --task action --source ../test_data/fun_theory.mp4 --save_folder ../output --debug_track
pycuda or tensorrt not installed.
[INFO] Loading pytorch trtpose model with "../weights\pose_estimation\trtpose\densenet121_baseline_att_256x256_B_epoch_160.pth"
[INFO] Loading pytorch reid model: ../weights\tracker\deepsort\wideresnet_mars.pth.
[INFO] Writing output to ../output\fun_theory_trtpose_deepsort_mars_wideresnet_dnn_torch.avi
[INFO] Writing output to ../output\fun_theory_trtpose_deepsort_mars_wideresnet_dnn_torch_debug.avi
Run | fun_theory.mp4 ---------------------------------------- 0% -:--:-- 66.85fps[INFO] Saving video to : ../output\fun_theory_trtpose_deepsort_mars_wideresnet_dnn_torch.avi
Run | fun_theory.mp4 ---- ----------------------------------- 11% -:--:-- 0.03fps
Traceback (most recent call last):
File "demo.py", line 138, in
main()
File "demo.py", line 98, in main
predictions = action_classifier.classify(predictions)
File "C:\Users\Utkarsh.Singh\Documents\Utkarsh\PoseTracking\human-action-recognition\src\lib\classifier\dnn\classifier.py", line 70, in classify
self.dict_id2clf[id] = self._create_classifier(id)
File "C:\Users\Utkarsh.Singh\Documents\Utkarsh\PoseTracking\human-action-recognition\src\lib\classifier\dnn\classifier.py", line 53, in
model_path, classes, window_size, human_id, threshold=threshold)
File "C:\Users\Utkarsh.Singh\Documents\Utkarsh\PoseTracking\human-action-recognition\src\lib\classifier\dnn\classifier.py", line 172, in init
self.model = pickle.load(f)
ModuleNotFoundError: No module named 'sklearn.neighbors._classification'

I try to run demo.py script and i've got this error message

Traceback (most recent call last):
File "demo.py", line 171, in
main()
File "demo.py", line 65, in main
tracker = get_tracker(**tracker_kwargs)
File "D:\AI_POSE\human-action-recognition\src\lib\tracker_init_.py", line 11, in get_tracker
return trackersname
File "D:\AI_POSE\human-action-recognition\src\lib\tracker\deepsort\deepsort.py", line 15, in init
self.extractor = FeatureExtractor(**kwargs)
File "D:\AI_POSE\human-action-recognition\src\lib\tracker\deepsort\reid_feature_extractor.py", line 43, in init
self.extractor = self._load_torch_model(model_path)
File "D:\AI_POSE\human-action-recognition\src\lib\tracker\deepsort\reid_feature_extractor.py", line 63, in _load_torch_model
state_dict = torch.load(model_path, map_location='cpu')['net_dict']
File "C:\Users\Admin\anaconda3\envs\action-recognition\lib\site-packages\torch\serialization.py", line 386, in load
return _load(f, map_location, pickle_module, **pickle_load_args)
File "C:\Users\Admin\anaconda3\envs\action-recognition\lib\site-packages\torch\serialization.py", line 559, in _load
raise RuntimeError("{} is a zip archive (did you mean to use torch.jit.load()?)".format(f.name))
RuntimeError: D:/AI_POSE/human-action-recognition/weights\tracker\deepsort\siamese_mars.pth is a zip archive (did you mean to use torch.jit.load()?)

I am using window 10

How to run in CUDA 11?

how to import trt_pose module

Hi sir, Thanks for the wonderful project. Can you give me a step-by-step on how to build this project on Windows 10? Thanks in advance.

Is there a way to remove false positives while detecting the actions?
or any other helpful parameters to tweak.

Thank You

Thank you for your code, it's helpful for me, you are very nice person. but here is a question needing your help! T^T~
No error when running 'demo.py', but the terminal just stopped at this step for 40minutes without ending:

[INFO] Loading pytorch trtpose model with "../weights/pose_estimation/trtpose/densenet121_baseline_att_256x256_B_epoch_160.pth"
[INFO] Loading TensorRT reid model: ../weights/tracker/deepsort/siamese_market1501.trt.
[INFO] Writing output to ../output/fun_theory_trtpose_deepsort_market1501_siamesenet.avi
[INFO] Writing output to ../output/fun_theory_trtpose_deepsort_market1501_siamesenet_debug.avi
Run | fun_theory.mp4 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0% -:--:-- 76.87fps[INFO] Saving video to : ../output/fun_theory_trtpose_deepsort_market1501_siamesenet.avi

Hope your reply~~~~