OVMR: Open-Vocabulary Recognition with Multi-Modal References

Benchmark Model Rank Results
open-vocabulary-object-detection-on-lvis-v1-0OVMR#6AP novel-LVIS base training: 34.4