This repository contains a Python script (onnx_inference.py) for extracting and comparing image features using the DINOv2 model in ONNX format. The script uses the onnxruntime library to load and run the model.

• ONNXRuntime
• OpenCV
• NumPy

Models can be obtained by exporting PyTorch models to ONNX models. To export models to ONNX, you can utilize the onnx_export_merge branch.

In addition, you can download pre-exported models from:

Google Drive

• dinov2_vits14
• dinov2_vitb14

Hugging Face

• dinov2_vits14

1. Clone the repository:

   git clone [email protected]:sefaburakokcu/dinov2_onnx.git
   cd dinov2_onnx

2. Install the required Python dependencies:

   pip install -r requirements.txt

   Make sure you have Python and pip installed.

3. Place your input images in the inputs directory.

4. Run the inference script:

   python onnx_inference.py

   Optional arguments:

   • --onnx_model: Path to the ONNX model file (default: ./dinov2_vits14.onnx).
   • --image_folder: Path to the folder containing input images (default: ./inputs/).

This project is licensed under the MIT License.