/pytorch-softtriple

SoftTriple (ICCV2019) in pytorch

Primary LanguagePythonMIT LicenseMIT

SoftTriple

This is an unofficial implementation of "SoftTriple Loss: Deep Metric Learning Without Triplet Sampling" (ICCV 2019) in Pytorch.

Installation

cd pytorch-hdml
pip install pipenv
pipenv install

Download dataset

cd data
python cars196_downloader.py
python cars196_converter.py

Train CARS196 dataset

Execute a training script. When executed, the tensorboard log is saved.

pipenv shell
python train_softtriple.py

Result

CARS196 result on training(99 classes, 30000 iterations)

Loss

loss

t-SNE

train_tsne