Object Detection on FLIR Aligned
55.8APInfraNet-RGB-IR
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| InfraNet-RGB-IRMode=RGB+IR, Backbone=YOLOv12, Parameters=56.8M, Inference time (ms)=13.22026.07 | 55.8 | 89.3 | — | 84.3 | 81.7 | 83 | |
| InfraNet-RGB-IRMode=RGB+IR, Backbone=YOLOv8, Parameters=54.3M, Inference time (ms)=13.42026.07 | 55.2 | 89.4 | — | 85.8 | 79.9 | 82.8 | |
| InfraNet-RGB-IRMode=RGB+IR, Backbone=YOLOv5, Parameters=52.2M, Inference time (ms)=13.22026.07 | 55.1 | 89.5 | — | 85.1 | 81.8 | 83.3 | |
| InfraNet-RGB-IRMode=RGB+IR, Backbone=ResNet50, Parameters=73.8M, Inference time (ms)=14.42026.07 | 54.3 | 88.6 | — | 83.5 | 80.4 | 81.9 | |
| InfraNet-IRMode=IR, Backbone=YOLOv8, Parameters=28.2M, Inference time (ms)=12.52026.07 | 54 | 88.5 | — | 84.6 | 80.9 | 82.7 | |
| InfraNet-IRMode=IR, Backbone=YOLOv5, Parameters=26.0M, Inference time (ms)=11.62026.07 | 53.8 | 88.1 | — | 84.9 | 80.4 | 82.6 | |
| InfraNet-IRMode=IR, Backbone=YOLOv12, Parameters=24.7M, Inference time (ms)=12.02026.07 | 53.2 | 87.7 | — | 85 | 79.5 | 82.2 | |
| InfraNet-IRMode=IR, Backbone=ResNet50, Parameters=42.1M, Inference time (ms)=9.42026.07 | 53 | 87.3 | — | 84.9 | 78.8 | 81.7 | |
| WaveMambaBackbone=YOLOv8, Params (M)=69.12026.05 | 48.1 | 88.4 | — | — | — | — | |
| WaveMambaMode=RGB+IR, Backbone=YOLOv8, Parameters=69.1M, Inference time (ms)=40.02026.07 | 48.1 | 88.4 | — | 84.2 | 80.9 | 82.5 | |
| WaveMambaMode=RGB+IR, Backbone=ResNet50, Parameters=193.2M, Inference time (ms)=53.22026.07 | 47.9 | 86.5 | — | 84.2 | 77.9 | 80.9 | |
| WaveMambaMode=RGB+IR, Backbone=YOLOv5, Parameters=45.6M, Inference time (ms)=44.12026.07 | 47.5 | 85.8 | — | 83.9 | 79.7 | 81.7 | |
| SAFER-DEIMBackbone=HGNetv2-B0, Params (M)=19.82026.05 | 46.5 | 85.8 | — | — | — | — | |
| TIRDetMode=IR, Backbone=CSPD532026.07 | 44.3 | 81.4 | — | — | — | — | |
| ESSFNBackbone=YOLOv8, Params (M)=80.22026.05 | 42.3 | 80.8 | — | — | — | — | |
| ESSFNMode=RGB+IR, Backbone=YOLOv8, Parameters=80.2M, Inference time (ms)=47.02026.07 | 42.3 | 80.8 | — | 81.4 | 73.5 | 77.2 | |
| CrossFormerBackbone=YOLOv5, Params (M)=340.02026.05 | 42.1 | 79.3 | — | — | — | — | |
| CrossFormerMode=RGB+IR, Backbone=YOLOv5, Parameters=340.0M, Inference time (ms)=80.02026.07 | 42.1 | 79.3 | — | 78.1 | 72.8 | 75.4 | |
| MFPTBackbone=ResNet50, Params (M)=200.02026.05 | 41.9 | 80 | — | — | — | — | |
| MFPTMode=RGB+IR, Backbone=ResNet50, Parameters=200.0M, Inference time (ms)=80.02026.07 | 41.9 | 80 | — | 79.3 | 72.9 | 76 | |
| G-DINOSetup=Fine-tuning (FT), Backbone=Swin-T2026.05 | 41.4 | 74.9 | 41.4 | — | — | — | |
| YOLOv8-IRMode=IR, Backbone=YOLOv8, Parameters=43.7M, Inference time (ms)=22.02026.07 | 38.3 | 72.9 | — | 75.1 | 65.3 | 69.9 | |
| Thermal-DetZero-shot=true, Backbone=Swin-T2026.05 | 37.2 | 66.4 | 35.9 | — | — | — | |
| Thermal-DetSetup=Zero-shot, Backbone=Swin-T, Pre-training=Synthetic GroundingCap-1M, M3FD, MMPD, MS22026.05 | 37.2 | 66.4 | 35.9 | — | — | — | |
| LLMDetVenue=CVPR’25, Zero-shot=true, Backbone=Swin-T2026.05 | 35.9 | 62.8 | 34.8 | — | — | — | |
| LLMDetSetup=Zero-shot, Backbone=Swin-T, Pre-training=GroundingCap-1M2026.05 | 35.9 | 62.8 | 34.8 | — | — | — | |
| MM-GDINOZero-shot=true, Backbone=Swin-T2026.05 | 35.4 | 61.9 | 37.1 | — | — | — | |
| MM-GDINOSetup=Zero-shot, Backbone=Swin-T, Pre-training=O365, GoldG, GRIT, V3Det2026.05 | 35.4 | 61.9 | 37.1 | — | — | — | |
| G-DINOVenue=ECCV’24, Zero-shot=true, Backbone=Swin-T2026.05 | 33.7 | 63.6 | 31.3 | — | — | — | |
| G-DINOSetup=Zero-shot, Backbone=Swin-T, Pre-training=O365, GoldG, Cap4M2026.05 | 33.7 | 63.6 | 31.3 | — | — | — | |
| YOLOv8-RGBMode=RGB, Backbone=YOLOv8, Parameters=43.7M, Inference time (ms)=22.02026.07 | 28.2 | 66.3 | — | 71.6 | 62.2 | 66.6 | |
| T-Rex2Venue=ECCV’24, Zero-shot=true, Backbone=Swin-T2026.05 | 27.6 | 51.4 | 25.5 | — | — | — | |
| T-Rex2Setup=Zero-shot, Backbone=Swin-T, Pre-training=10M data from various resources2026.05 | 27.6 | 51.4 | 25.5 | — | — | — | |
| YOLO-WorldVenue=CVPR’24, Zero-shot=true, Backbone=Swin-T2026.05 | 26.6 | 48.7 | 24.1 | — | — | — | |
| YOLO-WorldSetup=Zero-shot, Backbone=YOLOv8-L, Pre-training=O365, GoldG, CC3M2026.05 | 26.6 | 48.7 | 24.1 | — | — | — | |
| GLIPVenue=NeurIPS’22, Zero-shot=true, Backbone=Swin-T2026.05 | 25.1 | 47.1 | 22.6 | — | — | — | |
| GLIPSetup=Zero-shot, Backbone=Swin-T, Pre-training=O365, GoldG, Cap4M2026.05 | 25.1 | 47.1 | 22.6 | — | — | — |