# DEIM-X

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

- Family: DEIM (Intellindust AI Lab)
- Parameters: 62.0M
- GFLOPs: 202.0
- Default input size: 640px
- Detection: detr · nms-free
- Architecture: transformer · HGNetv2 backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 56.5%
- Paper: https://arxiv.org/abs/2412.04234
- Original code: https://github.com/Intellindust-AI-Lab/DEIM

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 59.6 | 3.0 | 328.8 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 59.6 | 2.9 | 349.8 | 399 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 59.6 | 14.1 | 70.9 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 59.6 | 8.0 | 125.1 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 59.6 | 39.8 | 25.1 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 59.6 | 15.0 | 66.7 | 403 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 59.5 | 70.7 | 14.1 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 59.6 | 58.6 | 17.1 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 59.6 | 0.4 | 2635.6 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

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

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