Implementation of XXXX 2023 paper Deep Metric Loss for Multimodal Learning.
This repository contains the code and the synthetic data to reproduce the result from the paper:
MultiModalLoss(num_classes, num_modalities, proxies_per_class=20, gamma=0.1)
Parameters:
- num_classes: The number of classes.
- num_modalities: The number of modalities.
- proxies_per_class: The number of proxies per class. The papaer uses 20.
- gamma: Scaling factor.
- Python (3.7.9)
- PyTorch (1.9.0)
- pytorch_metric_learning (0.9.99)
This code is inspired by SoftTriple Loss and pytorch-metric-learning
If you find the MultiModal Loss is userful, please cite the above paper:
@article{
}