FBK-HUPBA Submission to the EPIC-Kitchens 2019 Action Recognition Challenge
This is an incremental submission to a domain-specific challenge for action recognition in kitchen videos.
The authors tackled the EPIC-Kitchens 2019 action recognition challenge by developing CNN-LSTA and HF-TSN variants and using an ensemble, achieving top-1 accuracies of 35.54% on S1 and 20.25% on S2.
In this report we describe the technical details of our submission to the EPIC-Kitchens 2019 action recognition challenge. To participate in the challenge we have developed a number of CNN-LSTA [3] and HF-TSN [2] variants, and submitted predictions from an ensemble compiled out of these two model families. Our submission, visible on the public leaderboard with team name FBK-HUPBA, achieved a top-1 action recognition accuracy of 35.54% on S1 setting, and 20.25% on S2 setting.