Closed shubhamjha97 closed 5 years ago
@shubhamjha97 The train/test splits are the same as the original UCF and HMDB splits. For evaluation, it's just simple classification accuracy. The pipeline follows the standard evaluation protocal: given a video, sample 25 frames, 10 crops (four corners+center&flip) for each frame
The test data has not been included in your project page. Can you please provide links to the test data matrices for the UCF and HMDB datasets?