https://github.com/sunsmarterjie/yolov12
YOLOv12: Attention-Centric Real-Time Object Detectors
https://github.com/sunsmarterjie/yolov12
Last synced: about 2 months ago
JSON representation
YOLOv12: Attention-Centric Real-Time Object Detectors
- Host: GitHub
- URL: https://github.com/sunsmarterjie/yolov12
- Owner: sunsmarterjie
- License: agpl-3.0
- Created: 2025-02-06T20:05:57.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2025-02-17T02:33:16.000Z (about 2 months ago)
- Last Synced: 2025-02-17T02:36:20.602Z (about 2 months ago)
- Language: Python
- Size: 1.17 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-yolo - **Yolov12** - Centric Real-Time Object Detectors'](https://arxiv.org/abs/2502.12524) YOLOv12-N achieves 40.6% mAP with an inference latency of 1.64 ms on a T4 GPU, outperforming advanced YOLOv10-N / YOLOv11-N by 2.1%/1.2% mAP with a comparable speed. (Uncategorized / Uncategorized)
- awesome-yolo-object-detection - YOLOv12 - Centric Real-Time Object Detectors". (**[arXiv 2025](https://arxiv.org/abs/2502.12524)**) (Summary)
README
YOLOv12
YOLOv12: Attention-Centric Real-Time Object Detectors
[Yunjie Tian](https://sunsmarterjie.github.io/)1, [Qixiang Ye](https://people.ucas.ac.cn/~qxye?language=en)2, [David Doermann](https://cse.buffalo.edu/~doermann/)1
1 University at Buffalo, SUNY, 2 University of Chinese Academy of Sciences.
Comparison with popular methods in terms of latency-accuracy (left) and FLOPs-accuracy (right) trade-offs
[](https://arxiv.org/abs/2502.12524) [](https://huggingface.co/spaces/sunsmarterjieleaf/yolov12) 
[](https://blog.roboflow.com/use-yolov12-with-roboflow/#deploy-yolov12-models-with-roboflow)
## Updates
- 2025/02/22: [YOLOv12 TensorRT CPP Inference Repo + Google Colab Notebook Support](https://github.com/mohamedsamirx/YOLOv12-TensorRT-CPP).
- 2025/02/22: [How to train YOLO12 on a custom dataset | Step-by-step guide](https://youtu.be/dO8k5rgXG0M) by [Noor](https://github.com/noorkhokhar99).
- 2025/02/22: [Android deploy](https://github.com/mpj1234/ncnn-yolov12-android/tree/main). [TensorRT-YOLO](https://github.com/laugh12321/TensorRT-YOLO) accelerates yolo12 inference. Thanks to them!
- 2025/02/21: Try yolo12 for classification, oriented bounding boxes, pose estimation, and instance segmentation at [ultralytics](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/cfg/models/12). Thanks to them!
- 2025/02/20: [Any computer or edge device?](https://github.com/roboflow/inference) Support yolo12 now.
- 2025/02/20: [ONNX CPP Version](https://github.com/mohamedsamirx/YOLOv12-ONNX-CPP). [Train a yolov12 model on a custom dataset?](https://blog.roboflow.com/train-yolov12-model/) An introduction at [Youtube](https://www.youtube.com/watch?v=1YZDsZL_VyI&t=120s).
- 2025/02/19: [arXiv version](https://arxiv.org/abs/2502.12524) is public. [Demo](https://huggingface.co/spaces/sunsmarterjieleaf/yolov12) is available (try [Demo2](https://huggingface.co/spaces/sunsmarterjieleaf/yolov12_demo2) [Demo3](https://huggingface.co/spaces/sunsmarterjieleaf/yolov12_demo3) if busy).
Abstract
Enhancing the network architecture of the YOLO framework has been crucial for a long time but has focused on CNN-based improvements despite the proven superiority of attention mechanisms in modeling capabilities. This is because attention-based models cannot match the speed of CNN-based models. This paper proposes an attention-centric YOLO framework, namely YOLOv12, that matches the speed of previous CNN-based ones while harnessing the performance benefits of attention mechanisms.
YOLOv12 surpasses all popular real-time object detectors in accuracy with competitive speed. For example, YOLOv12-N achieves 40.6% mAP with an inference latency of 1.64 ms on a T4 GPU, outperforming advanced YOLOv10-N / YOLOv11-N by 2.1%/1.2% mAP with a comparable speed. This advantage extends to other model scales. YOLOv12 also surpasses end-to-end real-time detectors that improve DETR, such as RT-DETR / RT-DETRv2: YOLOv12-S beats RT-DETR-R18 / RT-DETRv2-R18 while running 42% faster, using only 36% of the computation and 45% of the parameters.
## Main Results
| Model | size
(pixels) | mAPval
50-95 | Speed
T4 TensorRT10
| params
(M) | FLOPs
(G) |
| :----------------------------------------------------------------------------------- | :-------------------: | :-------------------:| :------------------------------:| :-----------------:| :---------------:|
| [YOLO12n](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12n.pt) | 640 | 40.6 | 1.64 | 2.6 | 6.5 |
| [YOLO12s](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12s.pt) | 640 | 48.0 | 2.61 | 9.3 | 21.4 |
| [YOLO12m](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12m.pt) | 640 | 52.5 | 4.86 | 20.2 | 67.5 |
| [YOLO12l](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12l.pt) | 640 | 53.7 | 6.77 | 26.4 | 88.9 |
| [YOLO12x](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12x.pt) | 640 | 55.2 | 11.79 | 59.1 | 199.0 |
## Installation
```
wget https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.3/flash_attn-2.7.3+cu11torch2.2cxx11abiFALSE-cp311-cp311-linux_x86_64.whl
conda create -n yolov12 python=3.11
conda activate yolov12
pip install -r requirements.txt
pip install -e .
```
## Validation
[`yolov12n`](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12n.pt)
[`yolov12s`](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12s.pt)
[`yolov12m`](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12m.pt)
[`yolov12l`](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12l.pt)
[`yolov12x`](https://github.com/sunsmarterjie/yolov12/releases/download/v1.0/yolov12x.pt)
```python
from ultralytics import YOLO
model = YOLO('yolov12{n/s/m/l/x}.pt')
model.val(data='coco.yaml', save_json=True)
```
## Training
```python
from ultralytics import YOLO
model = YOLO('yolov12n.yaml')
# Train the model
results = model.train(
data='coco.yaml',
epochs=600,
batch=256,
imgsz=640,
scale=0.5, # S:0.9; M:0.9; L:0.9; X:0.9
mosaic=1.0,
mixup=0.0, # S:0.05; M:0.15; L:0.15; X:0.2
copy_paste=0.1, # S:0.15; M:0.4; L:0.5; X:0.6
device="0,1,2,3",
)
# Evaluate model performance on the validation set
metrics = model.val()
# Perform object detection on an image
results = model("path/to/image.jpg")
results[0].show()
```
## Prediction
```python
from ultralytics import YOLO
model = YOLO('yolov12{n/s/m/l/x}.pt')
model.predict()
```
## Export
```python
from ultralytics import YOLO
model = YOLO('yolov12{n/s/m/l/x}.pt')
model.export(format="engine", half=True) # or format="onnx"
```
## Demo
```
python app.py
# Please visit http://127.0.0.1:7860
```
## Acknowledgement
The code is based on [ultralytics](https://github.com/ultralytics/ultralytics). Thanks for their excellent work!
## Citation
```BibTeX
@article{tian2025yolov12,
title={YOLOv12: Attention-Centric Real-Time Object Detectors},
author={Tian, Yunjie and Ye, Qixiang and Doermann, David},
journal={arXiv preprint arXiv:2502.12524},
year={2025}
}
```