The specific idea is to classify the images first, and then vectorize the classification boundaries. This process is very complicated. According to different images, there are many differences in parameter settings and processes ... you can't learn them quickly.
The landlord should find someone who can help you.