search

GuangmingZhu / Conv3D_BICLSTM

Learning Spatiotemporal Features using 3DCNN and Convolutional LSTM for Gesture Recognition
http://openaccess.thecvf.com/content_ICCV_2017_workshops/w44/html/Zhang_Learning_Spatiotemporal_Features_ICCV_2017_paper.html
MIT License
56 stars 14 forks source link

Training code #4

Open sumeetssaurav opened 5 years ago

sumeetssaurav commented 5 years ago

Hello,

Congrats, for the wonderful work. I am facing certain issues while training on the IsoGD database. This part of the code throwing index out range error.

tf.contrib.layers.l2_regularizer(weight_decay)(networks.all_params[50]) IndexError: list index out of range

I changed the index, but getting very poor accuracy. What could be the reason, could you plz help me in resolving the issue.

Thanking You

emily19941102 commented 5 years ago

I can generate RGB images,but cannot generate flow images.It seems that .brox is not used.Can you help me? THX! @sumeetssaurav

GuangmingZhu commented 5 years ago

You can obtain the RGB/Depth/Flow images of IsoGD from the link: https://pan.baidu.com/s/12qC5sVhQAPizpsr0axVehw Password: fymm

GuangmingZhu commented 5 years ago

@sumeetssaurav This work is done with the old version of Tensorlayer. You can refer to the new version using keras in TF-1.2: https://github.com/GuangmingZhu/ContinuousGR and https://github.com/GuangmingZhu/AttentioninConvLSTM.