https://github.com/neka-nat/pytorch-softtriple

SoftTriple (ICCV2019) in pytorch
https://github.com/neka-nat/pytorch-softtriple

deep-learning deep-metric-learning iccv2019 pytorch

Last synced: 6 months ago
JSON representation

SoftTriple (ICCV2019) in pytorch

• Host: GitHub
• URL: https://github.com/neka-nat/pytorch-softtriple
• Owner: neka-nat
• License: mit
• Created: 2019-10-08T12:26:55.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-10-11T00:01:25.000Z (almost 6 years ago)
• Last Synced: 2025-04-13T08:12:08.381Z (6 months ago)
• Topics: deep-learning, deep-metric-learning, iccv2019, pytorch
• Language: Python
• Homepage:
• Size: 11.1 MB
• Stars: 8
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# SoftTriple

This is an unofficial implementation of ["SoftTriple Loss: Deep Metric Learning Without Triplet Sampling" (ICCV 2019)](https://arxiv.org/abs/1909.05235) 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](assets/loss.png)

#### t-SNE

![train_tsne](assets/train_tsne.gif)