BMVC20 - Deep Metric Learning Meets DeepClustering: An Novel Unsupervised Approach for Feature Embedding
AIOZ - BMVC 2020: "Deep Metric Learning Meets DeepClustering: An Novel UnsupervisedApproach for Feature Embedding"
This repository is the implementation of CBSwR for Unsupervised Deep Metric Learning task. Our model achieved 55.9, 37.6 on NMI, and 47.5, 45.6 on R@1 over CUB200-2001 and Car196 datasets, respectively. For the detail, please refer to link.
Summary
- The proposed framework
- Prerequisites
- Preprocessing
- Training
- Testing
- Citation
- License
- More information
The proposed framework
Prerequisites
Python3
Please install dependence package by run following command:
pip install -r requirements.txt
Preprocessing
CUB200-2011
The CUB200-2011 dataset should be downloaded via link. The downloaded file should be extracted to data/cub200 directory.
To pre-process CUB200-2011 dataset, please follow:
$ python pre_process/pre_process_cub200.py
Car196
The Car196 dataset should be downloaded via img_link and ano_link. The downloaded files should be extracted to data/car196 directory.
To pre-process Car196 dataset, please follow:
$ python pre_process/pre_process_car196.py
Training
To train CBSwR model on CUB200-2011 dataset, please follow:
$ python train.py --dataset cub200 --model_name CBSwR_CUB200
To train CBSwR model on Car196 dataset, please follow:
$ python train.py --dataset car196 --model_name CBSwR_Car196
The training scores will be printed every epoch.
Testing
In this repo, we include the pre-trained weight of CBSwR_CUB200 and CBSwR_Car196 models.
For CBSwR_CUB200 pretrained model. Please download the link and move to new_checkpoint. The trained CBSwR_CUB200 model can be tested in CUB200 validation set via:
$ python test.py --dataset cub200 --checkpoint_path new_checkpoint/CBSwR_CUB200.pth
For CBSwR_Car196 pretrained model. Please download the link and move to new_checkpoint. The trained CBSwR_Car196 model can be tested in Car196 validation set via:
$ python test.py --dataset car196 --checkpoint_path new_checkpoint/CBSwR_Car196.pth
Citation
If you use this code as part of any published research, we'd really appreciate it if you could cite the following paper:
@inproceedings{aioz_cbswr_bmvc2020,
author = {Nguyen, Binh X and Nguyen, Binh D and Carneiro, Gustavo and Tjiputra, Erman and Tran, Quang D and Do, Thanh-Toan},
title = {Deep Metric Learning Meets Deep Clustering: An Novel Unsupervised Approach for Feature Embedding},
booktitle = {BMVC},
year = {2020}}
}
License
MIT License
More information
AIOZ AI Homepage: https://ai.aioz.io
AIOZ Network: https://aioz.network