# YOLO-NAS-M

> YOLO-NAS-M: one-stage object detector from the YOLO-NAS family. Available in LibreYOLO, the MIT-licensed open-source library (free for commercial use).

- Family: YOLO-NAS (Deci AI (acquired by NVIDIA))
- Parameters: 51.1M
- GFLOPs: 88.9
- Default input size: 640px
- Detection: anchor-free · nms
- Architecture: one-stage · Neural-Architecture-Search (AutoNAC) backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 51.55%
- Paper: https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md
- Original code: https://github.com/Deci-AI/super-gradients

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA A100 | PyTorch FP32 | 50.5 | 17.0 | 58.9 | 363 |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreYOLONASm.pt")
result = model("image.jpg", conf=0.25, iou=0.45)
```

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