Study/개인 연구
DataSet
SooHyun2i
2022. 5. 26. 23:18
Self-supervised learning 기반 Video Representation Learning
training dataset : UCF/HMDB(small-scale), Kinetics 시리즈(medium-scale), AudioSet(large-scale)
Action recognition
- UCF 101 : pBYOL(A Large-Scale Study on Unsupervised Spatiotemporal Representation Learning,CVPR2021), VideoMAE(VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training, 2021)
Masked autoencoders are scalable vision learners(FAIR 2021) 과 유사 -> Image 이고 위에는 video 버전
- HMDB51 : pBYOL, VideoMAE
Audio Classification
- ESC-50 : Broaden Your Views for Self-Supervised Video Learning(ICCV 2021)
Video retrieval
- UCF 101, HMDB51 : Self-supervised Video Representation Learning with Cross-Stream Prototypical Contrasting(WACV 2022)