Tag
Falcon Perception is a 0.6B-parameter early-fusion Transformer model released by TII UAE for open-vocabulary grounding and segmentation from natural language prompts, utilizing hybrid attention and specialized heads.
Recognize Anything Model (RAM) is a strong image tagging model with zero-shot generalization, now combined with Grounded-Segment-Anything for open-set object detection and segmentation, significantly outperforming CLIP and BLIP.