# RT-DETR-R34

> RT-DETR-R34: 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: 31.0M
- GFLOPs: 91.0
- Default input size: 640px
- Detection: detr · nms-free
- Architecture: transformer · ResNet-34 backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 48.5%
- 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 | 48.2 | 27.8 | 36.0 | 198 |
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 52.2 | 7.6 | 131.7 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 52.2 | 8.0 | 125.8 | 198 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 52.3 | 31.9 | 31.4 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 52.2 | 17.1 | 58.5 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 52.2 | 77.7 | 12.9 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 52.2 | 40.2 | 24.9 | 213 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 52.3 | 108.4 | 9.2 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 52.2 | 98.1 | 10.2 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreRTDETRr34.pt")
result = model("image.jpg", conf=0.25, iou=0.45)
```

Source: https://www.visionanalysis.org/model/rtdetr-r34. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).