DetCLIPv2: Scalable Open-Vocabulary Object Detection Pre-training via Word-Region Alignment

Benchmark Model Rank Results
object-detection-on-odinw-full-shot-13-tasksDetCLIPv2AP: 70.4