# DEIMv2-Atto

> DEIMv2-Atto: transformer object detector from the DEIMv2 family. Available in LibreYOLO, the MIT-licensed open-source library (free for commercial use).

- Family: DEIMv2 (Intellindust AI Lab)
- Parameters: 0.5M
- GFLOPs: 0.8
- Default input size: 320px
- Detection: detr · nms-free
- Architecture: transformer · DINOv3-distilled ViT backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 23.8%
- Paper: https://arxiv.org/abs/2509.20787
- Original code: https://github.com/Intellindust-AI-Lab/DEIMv2

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 27.5 | 43.9 | 22.8 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 27.5 | 15.4 | 64.9 | 20 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 27.5 | 89.1 | 11.2 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 27.5 | 81.4 | 12.3 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 27.5 | 102.0 | 9.8 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 27.5 | 40.4 | 24.7 | 24 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 25.8 | 132.1 | 7.6 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 27.5 | 137.9 | 7.3 | - |
| Raspberry Pi 5 | ONNX Runtime FP32 | 27.5 | 30.3 | 33.0 | - |
| Raspberry Pi 5 | PyTorch FP32 | 27.5 | 12.7 | 79.1 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreDEIMv2atto.pt")
result = model("image.jpg", conf=0.25, iou=0.45)
```

Source: https://www.visionanalysis.org/model/deimv2-atto. Benchmarks produced with LibreYOLO (https://github.com/Libre-YOLO/libreyolo).