Skip to content
/ OADis Public
forked from nirat1606/OADis

Official code for "Disentangling Visual Embeddings for Attributes and Objects" Published at CVPR 2022

Notifications You must be signed in to change notification settings

wqshmzh/OADis

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Disentangling Visual Embeddings for Attributes and Objects (OADis)

This repository provides dataset splits and code for Paper:

Disentangling Visual Embeddings for Attributes and Objects, CVPR 2022

Nirat Saini, Khoi Pham, Abhinav Shrivastava

VAW-CZSL Dataset

We provide compositional splits for Generalized CZSL, following prior works: The dataset and splits can be downloaded from: VAW-CZSL. This folder has a jupyter notebook vaw_dataset_orig.ipynb, and folder named compositional-split-natural. The folder also has metadata file which splits image ids for each split.

  • compositional-split-natural: lists attribute-object pairs for each split [training, validation and testing]. Images folfder has all relevant images used in VAW-CZSL dataset.
  • vaw_dataset_orig.ipynb explains the steps for creation of splits, more details can also be found in Supplementary material. This file build the dataset splits from scratch.

For building split files and metedata files from scratch, you need

  1. The VAW-dataset from the website: VAW.
  2. Some images are part of Visual Genome, and can be downloaded from the official website.

Code Instructions:

Pre-requisites:

  • Update the path for dataset images and log file in the config/*.yml files.
  • Download and dump the pre-trained models from here to a folder named saved_models

To run OADis for MIT-States Dataset:

Training:
python train.py --cfg config/mit-states.yml

Testing:
python test.py --cfg config/mit-states.yml --load mit_final.pth

Similar instructions can be used for other datasets: UT-Zappos and VAW-CZSL. The code works well, and is tested for:

Pytorch - 1.6.0+cu92
Python - 3.6.12
tensorboardx - v2.4

For more qualitative results and details, refer to the Project Page

For questions and queries, feel free to reach out to Nirat.

Citation

Please cite our CVPR 2022 paper if you use the this repo for OADis.

@InProceedings{Saini_2022_CVPR,
    author    = {Saini, Nirat and Pham, Khoi and Shrivastava, Abhinav},
    title     = {Disentangling Visual Embeddings for Attributes and Objects},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {13658-13667}
}

About

Official code for "Disentangling Visual Embeddings for Attributes and Objects" Published at CVPR 2022

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%