If the file is not related to AI research, "VicaTs" is also the name of a contributor on Shutterstock , who provides royalty-free stock footage, including: Caribbean snorkeling and underwater scenes. Mediterranean beach landscapes and resort parks. Food-related content like muffins and cheesecakes. Important Note
: The dataset includes 20,416 annotated videos .
The ViCaS dataset is designed to help AI models understand video content at both a broad (holistic) and detailed (pixel-level) scale. It is significantly larger than previous benchmarks like YouTube-VOS.