FBK-HUPBA 参加 EPIC-Kitchens 2019 动作识别挑战赛的方案
计算机视觉与模式识别
2019-06-24 v1
摘要
在本报告中,我们描述了提交给 EPIC-Kitchens 2019 动作识别挑战赛的技术方案。为参与该挑战赛,我们开发了若干 CNN-LSTA [3] 与 HF-TSN [2] 的变体,并提交了由这两类模型集成所得的预测。我们的提交(在公开排行榜上以队伍名 FBK-HUPBA 可见)在 S1 设定下取得了 35.54% 的 top-1 动作识别准确率,在 S2 设定下取得了 20.25%。
引用
@article{arxiv.1906.08960,
title = {FBK-HUPBA Submission to the EPIC-Kitchens 2019 Action Recognition Challenge},
author = {Swathikiran Sudhakaran and Sergio Escalera and Oswald Lanz},
journal= {arXiv preprint arXiv:1906.08960},
year = {2019}
}
备注
Ranked 3rd in the EPIC-Kitchens 2019 action recognition challenge, held as part of CVPR 2019