search

LZQthePlane / Online-Realtime-Action-Recognition-based-on-OpenPose

A skeleton-based real-time online action recognition project, classifying and recognizing base on framewise joints, which can be used for safety surveilence.
Apache License 2.0
661 stars 231 forks source link

how to train action recognization? #92

Open Z0Victor opened 7 months ago

Z0Victor commented 7 months ago

Thanks to your exciting work :), But i got few questions about your action recognize model:

  1. how do you prepare your dataset, and how to train it?
  2. Is the dataset that you use not large enough? I runned this code to predict a scene of a man fall down with people around, but the result showed that it is a high rate to miss or wrong recognize
Z0Victor commented 7 months ago

I know how to train it right now, but I used your .csv file without modify any of your code and faced confuse problem. After training, I can't load the weight file I have trained, the error print: model = load_model('framewise_recognition_trained_myself.h5') File "D:\Anaconda3\lib\site-packages\keras\engine\saving.py", line 264, in load_model load_weights_from_hdf5_group(f['model_weights'], model.layers) File "D:\Anaconda3\lib\site-packages\keras\engine\saving.py", line 901, in load_weights_from_hdf5_group str(len(filtered_layers)) + ' layers.') ValueError: You are trying to load a weight file containing 7 layers into a model with 0 layers.

DKW :( ,could you tell me.