Skip to content
PyTorch Implementation for SoftTriple Loss
Python
Branch: master
Clone or download
qian-qi Update SoftTriple.py
change weight from uint8 to bool for PyTorch 1.2
Latest commit 2417374 Sep 26, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
loss Update SoftTriple.py Sep 26, 2019
net Add files via upload Sep 20, 2019
LICENSE Initial commit Sep 20, 2019
README.md Add files via upload Sep 20, 2019
cars_annos.txt Add files via upload Sep 20, 2019
evaluation.py Add files via upload Sep 20, 2019
genCars.py Add files via upload Sep 20, 2019
train.py Add files via upload Sep 20, 2019

README.md

SoftTriple Loss

PyTorch Implementation for ICCV'19: "SoftTriple Loss: Deep Metric Learning Without Triplet Sampling"

Usage: Train on Cars196

Here is an example of using this package.

  1. Obtain dataset
wget http://imagenet.stanford.edu/internal/car196/car_ims.tgz
tar -xf car_ims.tgz
  1. Generate train/test sets
python genCars.py
  1. Learn 64-dimensional embeddings
python train.py --gpu 0 --dim 64 -C 98 --freeze_BN [folder with train and test folders]

Requirements

  • Python 3.7
  • PyTorch 1.1
  • scikit-learn 0.20.1

Citation

If you use the package in your research, please cite the paper:

@inproceedings{qian2019striple,
  author    = {Qi Qian and
               Lei Shang and
               Baigui Sun and
               Juhua Hu and
               Hao Li and
               Rong Jin},
  title     = {SoftTriple Loss: Deep Metric Learning Without Triplet Sampling},
  booktitle = {{IEEE} International Conference on Computer Vision, {ICCV} 2019},
  year      = {2019}
}
You can’t perform that action at this time.