# RT-DETRv4-S

> RT-DETRv4-S: transformer object detector from the RT-DETRv4 family. Available in LibreYOLO, the MIT-licensed open-source library (free for commercial use).

- Family: RT-DETRv4 (Intellindust AI Lab)
- Parameters: 10.0M
- GFLOPs: 25.0
- Default input size: 640px
- Detection: detr · nms-free
- Architecture: transformer · HGNetv2 backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 49.7%
- Paper: https://arxiv.org/abs/2510.25257
- Original code: https://arxiv.org/abs/2510.25257

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 52.8 | 10.3 | 97.5 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 52.8 | 10.1 | 99.3 | 139 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 52.9 | 29.6 | 33.7 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 52.9 | 22.1 | 45.3 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 52.8 | 66.7 | 15.0 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 52.9 | 39.8 | 25.1 | 142 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 52.9 | 92.3 | 10.8 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 52.9 | 97.1 | 10.3 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 52.8 | 1.8 | 562.5 | - |
| Raspberry Pi 5 | PyTorch FP32 | 52.8 | 1.1 | 879.9 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreRTDETRv4s.pt")
result = model("image.jpg", conf=0.25, iou=0.45)
```

Source: https://www.visionanalysis.org/model/rtdetrv4-s. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).