# YOLOv5nu

> YOLOv5nu: one-stage object detector from the yolov5u family. Planned for LibreYOLO (not yet packaged).

- Family: yolov5u
- Parameters: 2.6M
- GFLOPs: 7.7
- Default input size: 640px
- Detection: anchor-free · nms
- Architecture: one-stage · CSPDarknet backbone
- Weight license: non-permissive
- Paper-reported mAP@50-95: 34.3%

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 33.8 | 7.9 | 127.1 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 33.8 | 26.1 | 38.3 | 31 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 33.8 | 56.3 | 17.8 | 14 |
| Raspberry Pi 5 | ncnn FP32 | 33.8 | 11.9 | 84.0 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 33.8 | 6.4 | 156.6 | - |
| Raspberry Pi 5 | PyTorch FP32 | 33.8 | 5.4 | 184.6 | - |


Source: https://www.visionanalysis.org/model/yolov5nu. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).