Object Detection on Four threshold-sweep benchmarks
47.73mAP (Open)LEVIRDetNet
Evaluation results
| Method | Links | |
|---|---|---|
| LEVIRDetNetFPS (imgs/s)=15.20, Latency (ms)=65.78, FLOPs (G)=216.0, GPU Hardware=RTX 4090, Batch size=12026.06 | 47.73 | |
| DEIMv2FPS (imgs/s)=18.55, Latency (ms)=53.91, FLOPs (G)=174, GPU Hardware=RTX 4090, Batch size=12026.06 | 26.93 | |
| LAE-DINOFPS (imgs/s)=9.00, Latency (ms)=111.10, FLOPs (G)=300, GPU Hardware=RTX 4090, Batch size=12026.06 | 19.6 | |
| SAM3FPS (imgs/s)=3.86, Latency (ms)=258.87, FLOPs (G)=5036.94, GPU Hardware=RTX 4090, Batch size=12026.06 | 17.2 |