Back to Course
Computer Vision Engineering
Module 8 of 8
8. Action Recognition
1. Video Classification
Classifying a single image is easy. Classifying "Swimming" requires Time. 3D CNNs (Conv3d) convolve over Height, Width, AND Time.
2. SlowFast Networks
Two pathways:
- Slow: High resolution, low frame rate (Spatial details).
- Fast: Low resolution, high frame rate (Motion details).