how to input videos into C3D-tensorflow

hx173149 / C3D-tensorflow

C3D is a modified version of BVLC tensorflow to support 3D ConvNets.

MIT License

587 stars 265 forks source link

how to input videos into C3D-tensorflow #58

Open Caroline1994 opened 6 years ago

Caroline1994 commented 6 years ago

I have got those files when I run python train_c3d_ucf101.py,who can tell me how should I to use them?

2018-05-11 10 24 24

and I have got this when I run python predict_c3d_ucf101.py,who can tell me what's means of it?

2018-05-11 10 25 06

How do we input videos into the network for behavior recognition?

Hi,@JulioEndara @frankgu @hx173149 @chuckcho @cc786537662 thank you very much

binhhoangtieu commented 6 years ago

According to predict_c3d_ucf101.py, from left to right for each row: true label, class prob for true label, predicted label, class prob for predicted label

If you want to use pre-trained, just change the line in predict_c3d_ucf101.py model_name = "./models/c3d_ucf_model-xxxx".

Best

Caroline1994 commented 6 years ago

@binhhoangtieu thank you Could you tell me how to use my own training results to predict the behavior of video in real time,not in the picture?

zeynepgokce commented 6 years ago

@binhhoangtieu I also tried to predict my trained model. But when i declare model name as you said above, e.g., model_name = "./models/c3d_ucf_model-4999" , it does not work.

tensorflow.python.framework.errors_impl.NotFoundError: Key var_name/bout not found in checkpoint
     [[Node: save/RestoreV2_10 = RestoreV2[dtypes=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/cpu:0"](_arg_save/Const_0_0, save/RestoreV2_10/tensor_names, save/RestoreV2_10/shape_and_slices)]]

I solved this problem with this. Please check this link to understand what the files(.meta , .index, .data ) after training means @Caroline1994