Object Detection on VisDrone-VID (val)
30.3mAP@0.5RT-DETR-L
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RT-DETR-LParams=32.0 M, GFLOPs=103.5, FPS=33.5, Input Resolution=640 × 640, Inference Platform=NVIDIA Jetson Orin Nano, Inference Precision=TensorRT FP162026.05 | 30.3 | 15.5 | |
| YOLOv8sParams=11.1 M, GFLOPs=28.5, FPS=43.7, Input Resolution=640 × 640, Inference Platform=NVIDIA Jetson Orin Nano, Inference Precision=TensorRT FP162026.05 | 28.5 | 13.6 | |
| Proposed Motion-Guided FrameworkParams=3.01 M, GFLOPs=8.1, FPS=72.1, Input Resolution=640 × 640, Inference Platform=NVIDIA Jetson Orin Nano, Inference Precision=TensorRT FP162026.05 | 27.4 | 12.1 | |
| YOLO11nParams=2.58 M, GFLOPs=6.3, FPS=68.5, Input Resolution=640 × 640, Inference Platform=NVIDIA Jetson Orin Nano, Inference Precision=TensorRT FP162026.05 | 26 | 12.2 | |
| YOLOv8n (Baseline)Params=3.01 M, GFLOPs=8.1, FPS=79.6, Input Resolution=640 × 640, Inference Platform=NVIDIA Jetson Orin Nano, Inference Precision=TensorRT FP162026.05 | 23.3 | 10.4 |