AdmReg

Pytorch code for the paper:

Adaptive Margin Diversity Regularizer for handling Data Imbalance in Zero-Shot SBIR
ECCV 2020 (Spotlight) Titir Dutta, Anurag Singh, and Soma Biswas

Dependencies:

• Linux (tested on Ubuntu 16.04)
• NVIDIA GPU + CUDA CuDNN
• 7z

Key Implementation Details:

• Weights for class-balanced loss
• Focal loss
• Assigning weights to different loss

Citation

If you find our work helpful in your research, please cite it as:

@inproceedings{dutta2020adaptive,
  title={Adaptive Margin Diversity Regularizer for Handling Data Imbalance in Zero-Shot SBIR},
  author={Dutta, Titir and Singh, Anurag and Biswas, Soma},
  booktitle={European Conference on Computer Vision},
  pages={349--364},
  year={2020},
  organization={Springer}
}