Semantic video segmentation
Semantic segmentation
The classic deep learning model for semantic segmentation is FCN…. add more introducation about FCN.
Semantic video segmentation
For videos, a naive method is to segment the image frame one by one. But this strategy has several drawbacks:
- Inefficiency.
- It doesn’t take advantage of the video spatio-temporal information.
The video spatio-temporal information can be utilized for calculating more stable segmentation results.
There are some works like Feature Space Optimization for Semantic Video Segmentation adopt CRF to impose extra constraints.