Syed Ashar Javed

We couldn't try out the LSTMs due to lack of enough GPU memory. If you wish to implement it, just follow the 2nd paper from the README. You need to...

@guoyang007 : Sorry for the late reply. One straightforward way which most papers also follow is to not train the whole network end-to-end. So, you train the CNN, freeze the...

You're confusing 20 channels with 20 frames. At time t, the current frame is sent into the spatial stream as it is. The optical flow for t to t+10 frames...

There's no video with

Sorry, but I think the trained models were lost during some issue we had with our version control. Reproducing it shouldn't be a problem though. Maybe the difference from the...

The repo is a re-implementation of the 2 stream CNN paper by Simonyan et al (https://papers.nips.cc/paper/5353-two-stream-convolutional-networks-for-action-recognition-in-videos.pdf). The paper trains the streams individually and does late fusion with the softmax through...