HanzhiC / yolo-world-onnx

ONNX models of YOLO-World (an open-vocabulary object detection).

yolo-world-onnx

Installation | Usage | License

ONNX Models

Download

Installation

Install dependencies and downloads the pre-trained model checkpoints:

make install

Usage

Inference with PyTorch

Run inference using the PyTorch pre-trained model:

./infer_pytorch.py
./infer_pytorch.py --class-names person,shoes
./infer_pytorch.py --image-file data/images/dogs.jpg \
                   --class-names dog,eye,nose,ear,tail \
                   --iou-threshold 0.5 \
                   --score-threshold 0.09

Export ONNX

Export YOLO-World and NMS models to ONNX format:

./export_onnx.py
./export_nms_onnx.py

Inference with ONNX

Run inference using the exported ONNX model:

./infer_onnx.py
./infer_onnx.py --class-names person,shoes
./infer_onnx.py --image-file data/images/dogs.jpg \
                --class-names dog,eye,nose,ear,tail \
                --iou-threshold 0.5 \
                --score-threshold 0.09

Acknowledgements

YOLO-World is an open-vocabulary object detection model published in CVPR2024.

Check out the following resources for more information: Paper, Code

License

GPLv3