# PicoDet-M

> PicoDet-M: one-stage object detector from the PicoDet family. Available in LibreYOLO, the MIT-licensed open-source library (free for commercial use).

- Family: PicoDet (Baidu PaddlePaddle)
- Parameters: 2.1M
- GFLOPs: 2.5
- Default input size: 416px
- Detection: anchor-free · nms
- Architecture: one-stage · ESNet backbone
- Weight license: Apache-2.0
- Paper-reported mAP@50-95: 34.3%
- Paper: https://arxiv.org/abs/2111.00902
- Original code: https://github.com/PaddlePaddle/PaddleDetection

## Benchmarks (COCO val2017)

| Hardware | Runtime | mAP@50-95 | FPS | Latency (ms) | VRAM (MB) |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 37.3 | 10.7 | 93.4 | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 37.9 | 12.5 | 80.3 | 26 |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 37.9 | 32.4 | 30.8 | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 37.9 | 29.6 | 33.7 | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 37.9 | 46.0 | 21.7 | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 37.9 | 45.1 | 22.2 | 38 |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 37.3 | 65.7 | 15.2 | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 36.0 | 53.5 | 18.7 | - |

## Usage with LibreYOLO

```python
from libreyolo import LibreYOLO

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

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