Semantic-aware Triplet Loss for Image Classification

This repository contains the implementation of our paper: Semantic-aware Triplet loss for Image Classification

Dataset

This repo contains the codes to run our method on CIFAR-10 and CIFAR-100 datasets. If you have not downloaded the two datasets, just start running, the datasets will be downloaded automatically.

Installation

Please install the following libraries

torchmeta 1.5.0
tensorboard 2.4.0
pytorch 1.4.0

Train

To train on CIFAR-100 with Glove as semantic source, just run

./scripts/r18_glove_c100.sh

Acknowledgement

Our code is built upon https://github.com/weiaicunzai/pytorch-cifar100 . Thanks for the code!