Skip to content

Official repository of "FashionFail: Addressing Failure Cases in Fashion Object Detection and Segmentation".

License

Notifications You must be signed in to change notification settings

HammerLabML/fashionfail

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

FashionFail

The official repository of "FashionFail: Addressing Failure Cases in Fashion Object Detection and Segmentation".

TL;DR: Generic badge Generic badge Generic badge arXiv

Install

Create a new environment named ff (short for FashionFail):

conda create -n ff python=3.9
conda activate ff

Then, clone the repository and install the package via:

git clone https://github.com/rizavelioglu/fashionfail.git
cd fashionfail
pip install -e .[dev]  # full development, e.g. training, inference, evaluation

Note: If you just want to construct the FashionFail dataset, you could install the package via the following instead, which avoids downloading heavy packages:

pip install -e .       # basic usage, i.e. downloading dataset

Usage

We offer a comprehensive explanation of each step, including dataset creation, training, and evaluation. You can find detailed documentation at references/README.md

FashionFail Dataset

The dataset (annotations and image URLs) is available at Generic badge. Execute the following to download the dataset (which downloads the annotation files and images for each train, val, and test splits):

python fashionfail/data/make_dataset.py \
--save_dir "dir/to/save" \  # [optional] default: "~/.cache/fashionfail/"

An optional argument --save_dir can be set to construct the dataset in the preferred directory, but it is not recommended to alter the default location, as the training script expects the data to be at a specific location.

FashionFail Models

Trained models are available at Generic badge. To run inference using models execute the following, which downloads the model if necessary:

python fashionfail/models/predict_models.py \
--model_name "facere" \                       # or "facere+"
--out_dir "<OUTPUT_DIR>" \                    # predictions are stored here
--image_dir "<IMAGES_DIR>"                    # image dir to run inference for

Alternatively, you can check out the demo in a browser at Generic badge and run inference using the user interface.

Training

Please refer detailed explanation of training both Facere and Facere+ models at references/03_training.md

And see the inference for Attribute Mask R-CNN and Fashionformer at references/04_inference.md


Project Structure

The following project/directory structure is adopted: Cookiecutter Data Science-v1 by DrivenData.

.
├── LICENSE
├── notebooks/          # Jupyter Notebooks
├── pyproject.toml      # The requirements file for reproducing the environment
├── README.md
├── references/         # Explanatory materials
├── src/
│   └── data/           # Scripts to download or construct dataset
│   └── features/       # Scripts to turn raw data into features
│   └── models/         # Scripts to train models, use trained models to make predictions
│   └── visualization/  # Scripts to create exploratory and results oriented visualizations

License

TL;DR: Not available for commercial use, unless the FULL source code is open-sourced!
This project is intended solely for academic research. No commercial benefits are derived from it.
The code, datasets, and models are published under the Server Side Public License (SSPL).

Citation

If you find this repository useful in your research, please consider giving a star ⭐ and a citation:

@inproceedings{velioglu2024fashionfail,
  author    = {Velioglu, Riza and Chan, Robin and Hammer, Barbara},
  title     = {FashionFail: Addressing Failure Cases in Fashion Object Detection and Segmentation},
  booktitle = {IJCNN},
  year      = {2024},
  doi       = {https://doi.org/ng59},
  eprint    = {2404.08582}
}

About

Official repository of "FashionFail: Addressing Failure Cases in Fashion Object Detection and Segmentation".

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%