This repository was archived by the owner on Jun 17, 2025. It is now read-only.
This repository was archived by the owner on Jun 17, 2025. It is now read-only.
Not able to reproduce reported results on Pytorch #134
Description
I am unable to reproduce the results for ir-CSN-152 and ip-CSN-152 pretrained on Sports1M ft Kinetics and IG-65M ft Kinetics. When tested on Kinetics 400 val, the Acc@1 and Acc@5 are around 40% and 65% respectively. Table 5 in the paper indicates 82% and 95%. Are there issues with the Pytorch models?