S2VT
S2VT copied to clipboard
test error
I have a big problem about test(). When I ran test(), there were no sentences been generated. And I found that video_feat.shape[1] was always 0 rather than 80(n_frame_step). How could I fix this?
Maybe you are using cv2 with anaconda which is not compiled with ffmpeg. Thats is why you are not able to extract frames and u have no features at all. So in the extract features file. You can use skvideo instead of cv2.videocapture.
Thank you for your reply. I can run successfully. Here is another question, I want to use GPU tensorflow because of the long training time. But it appeared a big problem:ResourceExhaustedError:OMM when allocating tensor with shape[3000,4000]. How could I fix this? Or maybe it is difficult which may cause many works. Please let me konw. Thank you very much!