# RT-DETRv2-R50m

> RT-DETRv2-R50m: transformer object detector from the RT-DETRv2 family. Available in LibreYOLO, the MIT-licensed open-source library (free for commercial use).

- Family: RT-DETRv2 (Baidu)
- Parameters: 36.0M
- GFLOPs: 100.0
- Default input size: 640px
- Detection: detr · nms-free
- Architecture: transformer · PResNet backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 51.9%
- Paper: https://arxiv.org/abs/2407.17140
- Original code: https://github.com/lyuwenyu/RT-DETR/tree/main/rtdetrv2_pytorch

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 54.8 | 5.4 | 184.3 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 54.8 | 5.1 | 196.8 | 298 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 54.8 | 29.7 | 33.6 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 54.8 | 15.8 | 63.3 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 54.8 | 79.6 | 12.6 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 54.8 | 39.9 | 25.1 | 302 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 54.7 | 113.0 | 8.9 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 54.8 | 103.3 | 9.7 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 54.8 | 0.8 | 1326.5 | - |
| Raspberry Pi 5 | PyTorch FP32 | 54.8 | 0.4 | 2810.3 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreRTDETRv2r50m.pt")
result = model("image.jpg", conf=0.25, iou=0.45)
```

Source: https://www.visionanalysis.org/model/rtdetrv2-r50m. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).