Ji Lin
Ji Lin
Hi, sorry for the late reply. I did not specially take care of the RGBDiff modality since it is not used. Have you figured out the problem yet? I think...
Hi, sorry about the waiting but I haven't got enough resources these days to take care of that. Maybe sometime in the future. Stay tuned!
Hi, the release code does have TVM part. One of the most usual problems is camera IO. You may want to reduce the image resolution from the camera. I set...
Hi, for online recognition, we need to use a uni-directional TSM, but we have not included the code. But it would be easy to modify by yourself (just change the...
For uni-directional TSM, we used multiple frames as input during training, while only one frame as input during testing.
Hi, did you resolve the issue with the above suggestion? Thanks!
Hi, could you please check if the bottleneck is on data loading or model traninig?
Hi, I would suggest storing data on an SSD to speed up data loading.
You need to first train an online TSM model on Something-Something dataset (using mobilenet v2 backbone) and then convert it to the online model checkpoint format.
Hi, I have not trained MobileNetV2 on UCF by myself. I would suggest you fine-tune from the Kinetics pre-trained weights since smaller models are generally more difficult to train.