# RT-DETR-R50m

> RT-DETR-R50m: transformer object detector from the RT-DETR family. Available in LibreYOLO, the MIT-licensed open-source library (free for commercial use).

- Family: RT-DETR (Baidu)
- Parameters: 36.6M
- GFLOPs: 0.0
- Default input size: 640px
- Detection: detr · nms-free
- Architecture: transformer · ResNet-50 backbone
- Weight license: Apache-2.0
- Paper: https://arxiv.org/abs/2304.08069
- Original code: https://github.com/lyuwenyu/RT-DETR

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA A100 | PyTorch FP32 | 50.8 | 25.1 | 39.8 | 298 |
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 53.8 | 5.4 | 184.5 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 53.8 | 5.0 | 198.3 | 298 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 53.8 | 29.1 | 34.3 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 53.8 | 15.7 | 63.6 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 53.8 | 76.4 | 13.1 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 53.8 | 35.3 | 28.4 | 302 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 53.9 | 110.1 | 9.1 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 53.9 | 102.8 | 9.7 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreRTDETRr50m.pt")
result = model("image.jpg", conf=0.25, iou=0.45)
```

Source: https://www.visionanalysis.org/model/rtdetr-r50m. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).