Yongyi Tang

I think @Seleucia is right that at testing time the decoder_input containing a large portion of ground-truth data is used for prediction. I also find that in https://github.com/una-dinosauria/human-motion-prediction/blob/1b39a7300192511b7387e8f3f393daf0aca9c0c4/src/translate.py#L225 , when...

It makes sense to me. Thanks. It seems that you did not report the validation error, so what should you correct in your paper? It looks fine to me now...

> This is the way I did it. > > ``` > def group_norm(x, G=32, eps=1e-5, scope='group_norm') : > with tf.variable_scope(scope) : > N, D, H, W, C = x.get_shape().as_list()...

@675492062 I wonder why this implementation use `g_x = tf.reshape(g, [batchsize,out_channels, -1]) g_x = tf.transpose(g_x, [0,2,1])` after computing g_x or theta_x. In my opinion, the output of slim.conv2d should be...

Sorry, I only successfully extracted feature from the original two-stream net.

> Ping! Any update on this? I found that the pretrained model work better with the flow images which are extracted after resizing the rgb frames. And I used OpenCV3.3...

> Actually, the code I used above works fine and produces good results on the example vid. But would still be nice to get pointers if this is missing something...