Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection

Benchmark Model Rank Results
zero-shot-object-detection-on-odinwGrounding DINO#2Average Score: 26.1
zero-shot-segmentation-on-segmentation-in-the-wildGrounded-SAM#2Mean AP: 46.0
object-detection-on-odinw-full-shot-13-tasksGrounding DINO#3AP: 70.9
zero-shot-object-detection-on-mscocoGrounding DINO-L (without COCO data)#3AP: 52.5
visual-grounding-on-refcoco-valGrounding DINO L (Swin-L, fine-tuned)#4Accuracy: 90.56
zero-shot-object-detection-on-lvis-v1-0-minivalGroundingDINO-L#13AP: 33.9
object-detection-on-coco-val2017Grounding DINO L (1.5x image size)#14box AP: 63.0
visual-grounding-on-refcoco-refcoco-refcocog-overallGrounding DINO L (Swin-L, fine-tuned)#16Accuracy: 86.13
object-detection-on-coco-test-devGrounding DINO#18box mAP: 63.0