README.md

Generative multiview inpainting for object removal in large indoor spaces

This repository provides codes including network architecture, input data generation and Anti-specificity Loss proposed in our paper.

1) Network Architecture

The generator and discriminator of proposing network will be provided in the python file 'model.py'.

2) Training input data generation

Example image, guideline and mask are prepared in the 'data' directory.
The image is a sample test image from Places2 dataset (Progressive Growing of GANs for Improved Quality, Stability, and Variation).

You will be able to run the code to generate sample training input data.

python input_data_generation.py

Following images will be saved : cropped ground truth image I_gt, guideline L_gt, color prior I_cp, no color mask M_cp, imperfect guideline ~L and no edge zone N.

3) Anti-specificity Loss

The source code for the anti-specificity loss will be also available in 'Loss.py'.

Citation

Please consider citing the corresponding publication if you use this work in an academic context:

@article{kim2021generative,
  title={Generative multiview inpainting for object removal in large indoor spaces},
  author={Kim, Joohyung and Hyeon, Janghun and Doh, Nakju},
  journal={International Journal of Advanced Robotic Systems},
  volume={18},
  number={2},
  year={2021},
  publisher={SAGE Publications Sage UK: London, England}
}