SODA10M introduction
It includes 10 million unlabeled images and 20,000 annotated images. The SODA10M dataset collects scenes from different cities under different weather conditions, time periods, and locations. Sunny and rainy, day and night, urban high-speed parks... More importantly, the coverage is very wide. The 10 million unlabeled images come from 32 cities, covering most of the country.
