
Researchers from Tsinghua University built Count Anything, an AI model that counts objects across diverse images like crowds, cars, cells, and bacterial colonies. It combines two approaches: bounding boxes for large objects and dots for small, dense ones. The system builds on Meta's SAM3 model with small adapter components.
Tap to vote and see what everyone thinks.
Summary by ByteBrief
Google's DiffusionGemma is 4x faster than its other Gemma models