Object Detection on LVIS zero-shot (mini val)
31.9AP (Common)YOLO-World-L (FP32)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| YOLO-World-L (FP32)Bits=FP32, Size(MB)=181.4, BOPs(T)=98.39, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 31.9 | 32.9 | 23.6 | 35.5 | |
| YOLO-World-X (FP32)Bits=FP32, Size(MB)=281.11, BOPs(T)=149.08, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 31.5 | 33.4 | 24.4 | 36.6 | |
| YOLO-World-M (FP32)Bits=FP32, Size(MB)=110.88, BOPs(T)=51.82, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 26.6 | 28.6 | 19.7 | 31.9 | |
| RTVPPre-trained Data=O365,GlodG, Zero-shot=true2026.06 | 23.4 | 26.8 | 19.5 | 30.7 | |
| RTVPPre-trained Data=O365, Zero-shot=true2026.06 | 22.4 | 24.1 | 17.3 | 27.9 | |
| YOLO-W-sPre-trained Data=O365,GlodG, Zero-shot=true2026.06 | 21.7 | 24.2 | 16.4 | 27.8 | |
| YOLO-WPre-trained Data=O365, Zero-shot=true2026.06 | 21.1 | 23.5 | 16.2 | 27 | |
| ViLDPre-trained Data=O365, Zero-shot=true2026.06 | 20 | — | — | 28.3 | |
| Grounding-Dino-TPre-trained Data=O365,GlodG, Zero-shot=true2026.06 | 19.6 | 25.6 | 14.4 | 32.2 | |
| GLIP-TPre-trained Data=O365,GlodG, Zero-shot=true2026.06 | 19.5 | 24.9 | 17.7 | 31 | |
| YOLO-World-L (CR-QAT)Bits=4-4-8, Size(MB)=24.32, BOPs(T)=8.18, Quantization Granularity=Ch-T-H, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 13.2 | 14.8 | 11.8 | 16.8 | |
| YOLO-World-M (CR-QAT)Bits=4-4-8, Size(MB)=15.11, BOPs(T)=7.10, Quantization Granularity=Ch-T-H, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 13 | 14.8 | 9.2 | 17.5 | |
| YOLO-World-X (CR-QAT)Bits=4-4-8, Size(MB)=37.19, BOPs(T)=9.32, Quantization Granularity=Ch-T-H, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 12.9 | 15 | 9.3 | 17.9 | |
| GLIP-TPre-trained Data=O365, Zero-shot=true2026.06 | 12.8 | 17.8 | 13.5 | 22.2 | |
| YOLO-World-M (QAT)Bits=4-4-8, Size(MB)=15.11, BOPs(T)=7.10, Quantization Granularity=Ch-T-H, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 10.2 | 12.3 | 8.2 | 14.9 | |
| YOLO-World-L (QAT)Bits=4-4-8, Size(MB)=24.32, BOPs(T)=8.18, Quantization Granularity=Ch-T-H, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 9.8 | 11.4 | 6.1 | 13.7 | |
| YOLO-World-X (QAT)Bits=4-4-8, Size(MB)=37.19, BOPs(T)=9.32, Quantization Granularity=Ch-T-H, Evaluation Protocol=LVIS fixed AP protocol2026.03 | 9.5 | 10.8 | 4.1 | 13.1 |