[Project Page]

This is the official PyTorch implementation of "Bones Can't Be Triangles: Accurate and Efficient Vertebrae Keypoint Estimation through Collaborative Error Revision (ECCV 2024)."

This code was developed using Python 3.8 on an Ubuntu 18.04 system.

1. Install Required Packages: Use pip to install the necessary Python packages from the requirements.txt file:

   pip install -r requirements.txt

2. Data Preparation:

   • Obtain the dataset: The AASCE dataset can be requested from this link.

   • Organize the dataset: Move the downloaded dataset to the following directory structure:

     codes/preprocess_data/AASCE_rawdata/boostnet_labeldata
     

   • Run preprocessing: Navigate to the preprocessing code directory and execute the preprocessing script:

     cd codes/preprocess_data/
     python preprocess_data.py
     cd ..

1. Training Your Own Model:

   To train your model, execute the following command:

   bash train_interactive_keypoint_model.sh
   python train_AASCE.py
   

2. Inference:

   Once the data is prepared, run the following command to perform inference with the pre-trained model:

   python evaluate_AASCE.py
   

If you find this work or code is helpful in your research, please cite:

@inproceedings{kim2024Bones,
  title={Bones Can't Be Triangles: Accurate and Efficient Vertebrae Keypoint Estimation through Collaborative Error Revision},
  author={Kim, Jinhee and Kim, Taesung and Choo, Jaegul},
  booktitle={European Conference on Computer Vision},
  year={2024},
}