Object Detection on FLIR (test)
2.39mAP50Le MuMo JEPA
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Le MuMo JEPASetting=Fine-tune2026.03 | 2.39 | — | 0.1288 | — | — | — | — | — | |
| Le MuMo JEPASetting=Waymo→FLIR2026.03 | 1.56 | — | 0.1022 | — | — | — | — | — | |
| MS-GDINOModel Variant=Large2025.12 | 0.878 | — | — | — | 0.513 | 0.505 | — | — | |
| DAMSDet2025.12 | 0.866 | — | — | — | 0.481 | 0.493 | — | — | |
| LGFDBackbone=CSPDarknet, Input Modality=Infrared-only2026.01 | 0.861 | — | — | — | — | 0.458 | — | — | |
| MS-GDINOModel Variant=Tiny2025.12 | 0.86 | — | — | — | 0.476 | 0.485 | — | — | |
| MS-GDINOModel Variant=Base2025.12 | 0.859 | — | — | — | 0.492 | 0.494 | — | — | |
| SLGNetModality=RGB+IR, Trainable Params=12.1M2026.01 | 0.858 | — | — | — | 0.423 | 0.451 | — | — | |
| MS-YOLOWModel Variant=Large2025.12 | 0.853 | — | — | — | 0.477 | 0.489 | — | — | |
| IGTBackbone=Swin-T-Tiny, Input Modality=Infrared-visible2026.01 | 0.85 | — | — | — | — | 0.436 | — | — | |
| YOLO-CIRBackbone=CSPDarknet, Input Modality=Infrared-only2026.01 | 0.849 | — | — | — | — | — | — | — | |
| Fusion-Mamba2025.12 | 0.849 | — | — | — | 0.459 | 0.47 | — | — | |
| YOLO-FusionBackbone=CSPDarknet, Input Modality=Infrared-visible2026.01 | 0.847 | — | — | — | — | — | — | — | |
| RSDetBackbone=ResNet50, Modality=RGB+IR2024.01 | 0.839 | — | — | — | — | 0.438 | — | — | |
| RSDetModality=RGB+IR2026.01 | 0.839 | — | — | — | 0.401 | 0.438 | — | — | |
| GMDETR2025.12 | 0.839 | — | — | — | 0.426 | 0.458 | — | — | |
| COFNetModality=RGB+IR, Trainable Params=90.2M2026.01 | 0.836 | — | — | — | 0.417 | 0.446 | — | — | |
| MS-YOLOWModel Variant=Medium2025.12 | 0.835 | — | — | — | 0.459 | 0.478 | — | — | |
| OursData Type=IR+RGB, Backbone=CSPDarknet532026.01 | 0.832 | — | — | — | — | 0.446 | — | — | |
| CAFF-DINOResolution=640x6402025.12 | 0.831 | — | — | — | 0.423 | 0.456 | — | — | |
| FD2-NetBackbone=CSPDarknet, Input Modality=Infrared-visible2026.01 | 0.829 | — | — | — | — | — | — | — | |
| RGBTInput Modality=Infrared-only2026.01 | 0.829 | — | — | — | — | — | — | — | |
| YOLO-ACNBackbone=CSPDarknet, Input Modality=Infrared-only2026.01 | 0.823 | — | — | — | — | — | — | — | |
| YOLOXCPCF2025.12 | 0.821 | — | — | — | 0.412 | 0.446 | — | — | |
| YOLOv5mBackbone=CSPDarknet, Input Modality=Infrared-only2026.01 | 0.819 | — | — | — | — | — | — | — | |
| CrossModalNetModality=RGB+IR, Trainable Params=92.8M2026.01 | 0.817 | — | — | — | 0.391 | 0.433 | — | — | |
| UniRGB-IRModality=RGB+IR, Trainable Params=8.9M2026.01 | 0.814 | — | — | — | 0.402 | 0.441 | — | — | |
| UniRGB-IRData Type=IR+RGB, Backbone=Transformer2026.01 | 0.814 | — | — | — | — | 0.441 | — | — | |
| FFMData Type=IR+RGB, Backbone=CSPDarknet532026.01 | 0.814 | — | — | — | — | 0.423 | — | — | |
| MFPTBackbone=ResNet50, Modality=RGB+IR2024.01 | 0.8 | — | — | — | — | — | — | — | |
| CrossFormerBackbone=ViT, Input Modality=Infrared-visible2026.01 | 0.793 | — | — | — | — | 0.421 | — | — | |
| CrossFormerData Type=IR+RGB, Backbone=CSPDarknet532026.01 | 0.793 | — | — | — | — | 0.421 | — | — | |
| Ours2023.08 | 0.792 | 0.669 | 0.89 | 0.816 | 36.9 | 41.4 | — | — | |
| ICAFusionBackbone=CSPDarkNet53, Modality=RGB+IR2024.01 | 0.792 | — | — | — | — | 0.414 | — | — | |
| CSAABackbone=ResNet50, Modality=RGB+IR2024.01 | 0.792 | — | — | — | — | 0.413 | — | — | |
| ICAFusionModality=RGB+IR, Trainable Params=120.0M2026.01 | 0.792 | — | — | — | 0.369 | 0.414 | — | — | |
| ICAFusionData Type=IR+RGB, Backbone=CSPDarknet532026.01 | 0.792 | — | — | — | — | 0.414 | — | — | |
| CSAAData Type=IR+RGB, Backbone=ResNet502026.01 | 0.792 | — | — | — | — | 0.413 | — | — | |
| ICAFusion2025.12 | 0.792 | — | — | — | 0.369 | 0.414 | — | — | |
| CFTBackbone=CFB, Input Modality=Infrared-visible2026.01 | 0.787 | — | — | — | — | 0.402 | — | — | |
| CFT2025.12 | 0.787 | — | — | — | 0.355 | 0.402 | — | — | |
| Ours*variant=asterisk and double-dagger notation2023.08 | 0.786 | 0.652 | 0.902 | 0.804 | 35.8 | 40.8 | — | — | |
| CFTvariant=double-dagger notation2023.08 | 0.783 | 0.614 | 0.895 | 0.841 | 35.5 | 40.2 | — | — | |
| SMG-YInput Modality=Infrared-only2026.01 | 0.77 | — | — | — | — | — | — | — | |
| YOLOFusionBackbone=CSPDarkNet53, Modality=RGB+IR2024.01 | 0.766 | — | — | — | — | 0.398 | — | — | |
| TINetBackbone=ResNet50, Modality=RGB+IR2024.01 | 0.761 | — | — | — | — | 0.365 | — | — | |
| FasterRCNNModality=IR, Trainable Params=68.5M2026.01 | 0.758 | — | — | — | 0.316 | 0.376 | — | — | |
| ProbEnBackbone=ResNet50, Modality=RGB+IR2024.01 | 0.755 | — | — | — | — | 0.379 | — | — | |
| MultiMAE-MTSetting=Fine-tune2026.03 | 0.75 | — | 0.0583 | — | — | — | — | — | |
| GAFFBackbone=ResNet18, Modality=RGB+IR2024.01 | 0.746 | — | — | — | — | 0.374 | — | — | |
| LGADetBackbone=ResNet50, Modality=RGB+IR2024.01 | 0.745 | — | — | — | — | — | — | — | |
| Faster R-CNNBackbone=ResNet50, Modality=IR2024.01 | 0.744 | — | — | — | — | 0.376 | — | — | |
| Faster-RCNNBackbone=ResNet50, Input Modality=Infrared-only2026.01 | 0.744 | — | — | — | — | 0.376 | — | — | |
| DDQ-DETRBackbone=ResNet50, Modality=IR2024.01 | 0.739 | — | — | — | — | 0.371 | — | — | |
| DDQ-DETRModality=IR, Trainable Params=244.6M2026.01 | 0.739 | — | — | — | 0.322 | 0.371 | — | — | |
| BU-LTTvariant=dagger notation2023.08 | 0.732 | 0.574 | 0.865 | 0.756 | — | — | — | — | |
| CAPTM_3Backbone=ResNet50, Modality=RGB+IR2024.01 | 0.732 | — | — | — | — | — | — | — | |
| BU-ATTvariant=dagger notation2023.08 | 0.731 | 0.561 | 0.87 | 0.761 | — | — | — | — | |
| GAFF2023.08 | 0.729 | — | — | — | 30.9 | 37.3 | — | — | |
| GAFFBackbone=ResNet18, Input Modality=Infrared-visible2026.01 | 0.729 | — | — | — | — | 0.375 | — | — | |
| YOLOV8Modality=IR, Trainable Params=76.7M2026.01 | 0.729 | — | — | — | 0.318 | 0.383 | — | — | |
| YOLOv8l-IRData Type=IR, Backbone=CSPDarknet532026.01 | 0.729 | — | — | — | — | 0.383 | — | — | |
| GAFFBackbone=VGG16, Modality=RGB+IR2024.01 | 0.727 | — | — | — | — | 0.373 | — | — | |
| GAFFBackbone=VGG16, Input Modality=Infrared-visible2026.01 | 0.727 | — | — | — | — | 0.373 | — | — | |
| CFR2023.08 | 0.724 | 0.5777 | 0.8491 | 0.7449 | — | — | — | — | |
| CFR_3Backbone=VGG16, Modality=RGB+IR2024.01 | 0.724 | — | — | — | — | — | — | — | |
| ICAFusionBackbone=ResNet50, Modality=RGB+IR2024.01 | 0.72 | — | — | — | — | — | — | — | |
| ImageBindSetting=Waymo→FLIR2026.03 | 0.72 | — | 0.0549 | — | — | — | — | — | |
| Halfway fusionBackbone=VGG16, Modality=RGB+IR2024.01 | 0.715 | — | — | — | — | 0.358 | — | — | |
| Cascade R-CNNBackbone=ResNet50, Modality=IR2024.01 | 0.71 | — | — | — | — | 0.347 | — | — | |
| ImageBindSetting=Fine-tune2026.03 | 0.71 | — | 0.0554 | — | — | — | — | — | |
| CMPDBackbone=ResNet50, Modality=RGB+IR2024.01 | 0.694 | — | — | — | — | — | — | — | |
| MultiMAE-MTSetting=Waymo→FLIR2026.03 | 0.69 | — | 0.0526 | — | — | — | — | — | |
| YOLOV8Modality=RGB, Trainable Params=76.7M2026.01 | 0.663 | — | — | — | 0.242 | 0.282 | — | — | |
| YOLOv8l-RGBData Type=RGB, Backbone=CSPDarknet532026.01 | 0.663 | — | — | — | — | 0.282 | — | — | |
| RetinaNetBackbone=ResNet50, Modality=IR2024.01 | 0.661 | — | — | — | — | 0.315 | — | — | |
| RetinaNetModality=IR, Trainable Params=43.0M2026.01 | 0.661 | — | — | — | 0.253 | 0.315 | — | — | |
| SSDBackbone=VGG16, Modality=IR2024.01 | 0.655 | — | — | — | — | 0.296 | — | — | |
| SSDBackbone=VGG16, Input Modality=Infrared-only2026.01 | 0.655 | — | — | — | — | 0.296 | — | — | |
| Faster R-CNNBackbone=ResNet50, Modality=RGB2024.01 | 0.649 | — | — | — | — | 0.289 | — | — | |
| DDQ-DETRBackbone=ResNet50, Modality=RGB2024.01 | 0.649 | — | — | — | — | 0.309 | — | — | |
| DDQ-DETRModality=RGB, Trainable Params=244.6M2026.01 | 0.649 | — | — | — | 0.245 | 0.309 | — | — | |
| MultiMAE-SSSetting=Fine-tune2026.03 | 0.63 | — | 0.0481 | — | — | — | — | — | |
| FasterRCNNModality=RGB, Trainable Params=68.5M2026.01 | 0.622 | — | — | — | 0.212 | 0.277 | — | — | |
| MMTOD-UNIT2023.08 | 0.615 | 0.4943 | 0.7072 | 0.6447 | — | — | — | — | |
| MMTOD-CG2023.08 | 0.614 | 0.5026 | 0.7063 | 0.6331 | — | — | — | — | |
| Cascade R-CNNBackbone=ResNet50, Modality=RGB2024.01 | 0.56 | — | — | — | — | 0.247 | — | — | |
| MultiMAE-SSSetting=Waymo→FLIR2026.03 | 0.54 | — | 0.0421 | — | — | — | — | — | |
| SSDBackbone=VGG16, Modality=RGB2024.01 | 0.522 | — | — | — | — | 0.218 | — | — | |
| RetinaNetBackbone=ResNet50, Modality=RGB2024.01 | 0.512 | — | — | — | — | 0.219 | — | — | |
| RetinaNetModality=RGB, Trainable Params=43.0M2026.01 | 0.512 | — | — | — | 0.152 | 0.219 | — | — | |
| MultiMAE-SSSetting=Scratch2026.03 | 0.51 | — | 0.0386 | — | — | — | — | — | |
| Le MuMo JEPASetting=Scratch2026.03 | 0.41 | — | 0.03 | — | — | — | — | — | |
| ImageBindSetting=Scratch2026.03 | 0.24 | — | 0.019 | — | — | — | — | — | |
| FALCON-SFOD2025.12 | — | 0.437 | 0.699 | 0.619 | — | 0.585 | — | — | |
| FALCON-SFOD2025.12 | — | 0.138 | 0.219 | 0.271 | — | 0.209 | — | — | |
| ROI-PackingNetwork=Faster RCNN-X101-FPN [9], Image Type=Infrared, Size (%)=1002025.12 | — | — | — | — | — | — | -10.86 | 0.6 | |
| ROI-PackingNetwork=Faster RCNN-X101-FPN [9], Image Type=Infrared, Size (%)=752025.12 | — | — | — | — | — | — | -18.1 | 2.17 | |
| ROI-PackingNetwork=Faster RCNN-X101-FPN [9], Image Type=Infrared, Size (%)=502025.12 | — | — | — | — | — | — | -18.77 | 2.52 | |
| ROI-PackingNetwork=Faster RCNN-X101-FPN [9], Image Type=Infrared, Size (%)=252025.12 | — | — | — | — | — | — | -7.21 | 0.47 |