# D-FINE-M

> D-FINE-M: transformer object detector from the D-FINE family. Available in LibreYOLO, the MIT-licensed open-source library (free for commercial use).

- Family: D-FINE (USTC)
- Parameters: 19.0M
- GFLOPs: 57.0
- Default input size: 640px
- Detection: detr · nms-free
- Architecture: transformer · HGNetv2 backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 52.3%
- Paper: https://arxiv.org/abs/2410.13842
- Original code: https://github.com/Peterande/D-FINE

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA A100 | PyTorch FP32 | 55.1 | 21.7 | 46.1 | 187 |
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 57.8 | 6.9 | 144.4 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 57.8 | 7.0 | 143.3 | 188 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 57.9 | 23.7 | 42.3 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 57.7 | 15.8 | 63.5 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 57.8 | 59.8 | 16.7 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 57.8 | 28.8 | 34.8 | 189 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 57.6 | 74.5 | 13.4 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 57.8 | 85.3 | 11.7 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

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

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