Work

Egocentric action recognition from noisy videos

Year: 2020
Type: dissertation
Abstract: Recognizing human actions in videos can make machines better interpret the behavior of human beings.As one of the basic fields of computer vision, action recognition has attracted significant research attention.In particular, with the recent advance of deep learning, there is a huge boost of action recognition performance.Deep learning models that takes as input video clips can reliably recognize the actions compared with traditional methods using hand-crafted features. List of Tables4.1 Action recognition performance comparison with baseline methods. .4.2 Quantitative comparison on cooperating with different backbones. . .4.3 Ablation study for different parts of our model on recognizing interrupted actions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4.4 Ablation study for different parts of our model on localizing useful frames in super noisy videos. . . . . . . . . . . . . . . . . . . . . . . .5.1 Quantitative comparison on using different attention mechanisms. . .
Author: Lijin Yang
Language: English
Open Access status: green
OpenAlex ID: W3014414335
Cites: 73
Cited by: 2
Related to: 10