# YOLOv5xu

> YOLOv5xu: one-stage object detector from the yolov5u family. Planned for LibreYOLO (not yet packaged).

- Family: yolov5u
- Parameters: 97.2M
- GFLOPs: 246.4
- Default input size: 640px
- Detection: anchor-free · nms
- Architecture: one-stage · CSPDarknet backbone
- Weight license: non-permissive
- Paper-reported mAP@50-95: 53.2%

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 52.3 | 0.5 | 2179.7 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 52.3 | 5.0 | 200.2 | 449 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 52.3 | 14.7 | 68.1 | 14 |
| Raspberry Pi 5 | ncnn FP32 | 52.3 | 1.0 | 977.7 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 52.3 | 0.3 | 2837.2 | - |
| Raspberry Pi 5 | PyTorch FP32 | 52.3 | 0.4 | 2484.4 | - |


Source: https://www.visionanalysis.org/model/yolov5xu. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).