# divide-and-conquer

**Repository Path**: zgpio/divide-and-conquer

## Basic Information

- **Project Name**: divide-and-conquer
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: LGPL-3.0
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2019-12-02
- **Last Updated**: 2020-12-21

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Deep Metric Learning

## Requirements

- Python >= 3.6.6
- SciPy = 1.2
- scikit-learn = 0.20
- PyTorch ([pytorch.org](http://pytorch.org)) = 1.0
- Faiss with GPU support ([Faiss](https://github.com/facebookresearch/faiss)) = 1.5
- Datasets
    - [CUB-200-2011](http://www.vision.caltech.edu/visipedia/CUB-200.html)
    - [In-Shop Clothes](http://mmlab.ie.cuhk.edu.hk/projects/DeepFashion/InShopRetrieval.html)
    - [Stanford Online Products](http://cvgl.stanford.edu/projects/lifted_struct/)
    - [Cars196](https://ai.stanford.edu/~jkrause/cars/car_dataset.html)

## Usage

You can see how to set up the experiment through `python experiment.py --help`.

### choose dataset
You can use the option `--dataset={dataset}` for the datasets.
It should be `inshop`/`sop`/`cub`/`car`.

### example
The following command will train the model with Margin loss on the In-Shop Clothes dataset for `200` epochs and
a batch size of `100` while splitting the embedding layer with `8` clusters and
finetuning the model from epoch `190` on.
```sh
CUDA_VISIBLE_DEVICES=0,1,2,3 python experiment.py --dataset=inshop \
--dir=dev --exp=0 --random-seed=0 --nb-clusters=8 --nb-epochs=200 \
--sz-batch=100 --backend=faiss-gpu  --embedding-lr=1e-5 --embedding-wd=1e-4 \
--backbone-lr=1e-5 --backbone-wd=1e-4 --finetune-epoch=190 --loss margin
```

### browse result

The model checkpoints and log files are saved in the selected log-directory specified by `--dir` option.
You can print a summary of the results with `python browse_results <log path>`.

## License
[LGPL](LICENSE)

## Reference
- [Divide and Conquer the Embedding Space](https://github.com/CompVis/metric-learning-divide-and-conquer.git)
- [Deep Metric Learning Baselines](https://github.com/Confusezius/Deep-Metric-Learning-Baselines)