In the field of computer vision, how to iterate existing visual models to ensure recognition effects and how to better deal with fragmented massive long-tail models have always been difficult problems that need to be overcome. Xing Xiaohuan will be broadcast through two Q&A issues