Huy V. Vo, Patrick Pérez, Jean Ponce - ECCV 2020

Video presentation: https://www.youtube.com/watch?v=v2vzEXOvUMs&feature=youtu.be

The code is written and tested with Matlab 2017a. Modifications might be necessary to run it with other versions of Matlab.

You need a GPU to extract features from pre-trained VGG19.

git clone https://github.com/huyvvo/rOSD.git

cd rOSD; mkdir data

Download VGG19 model and put it in data/.matconvnet/models.

Download the VOC_6x2 dataset here and put it in data. As a sanity check, you should have the file data/vocx/aeroplane_left/aeroplane_left.mat.

The main scripts for testing the code on VOC_6x2 are run_proposals.m, run_scores.m and run_rOSD.m. In matlab, from the rOSD folder, run

set_path; run_proposals; run_scores; run_rOSD;

For a quick test, you can use the precomputed scores instead. In matlab, from the rOSD folder, run

set_path; run_proposals;

Download the scores and put them in data/vocx_cnn by class, then from the rOSD folder, run

set_path; run_rOSD;

@INPROCEEDINGS{Vo20rOSD,
  title     = {Toward unsupervised, multi-object discovery in large-scale image collections},
  author    = {Vo, Huy V. and P{\'e}rez, Patrick and Ponce, Jean},
  booktitle = {Proceedings of the European Conference on Computer Vision ({ECCV})},
  year      = {2020}
}

This project is licensed under the MIT License - see the LICENSE file for details.

The code for Probabilistic Hough Matching (PHM) algorithm is taken from the project page of the paper "Unsupervised Object Discovery and Localization in the Wild".

The code for OSD and rOSD are modified from our previous project.

We use MatConvNet for running neural networks on Matlab.

This work was supported in part by the Inria/NYU collaboration, the Louis Vuitton/ENS chair on artificial intelligence and the French government under management of Agence Nationale de la Recherche as part of the “Investissements d’avenir” program, reference ANR19-P3IA-0001 (PRAIRIE 3IA Institute). Huy V. Vo was supported in part by a Valeo/Prairie CIFRE PhD Fellowship.