Train semantic segmentation model with unsupervise leanring


Is is possible to train a semantic segmentation model by unsupervised learning?

  1. Train the model with supervised learning
  2. Use trained model to generate labels image
  3. Put label images into training
  4. Repeat step 1~3

By the way, what kind of camera you would recommend for automobile application?
It need to achieve 30fps, with 720 resolution