This repository contains a Pytorch implementation of "Object level Visual Reasoning in Videos", F. Baradel, N. Neverova, C. Wolf, J. Mille, G. Mori, In ECCV 2018.
Links: Project page | Camera-ready | Masks-VLOG | Masks-EPIC-AR
We release code for training and testing our implementation. We encourage you to follow the steps below:
- preprocessing the video dataset
- rescaling an entire dataset (WxH=256x256 and fps=30)
- testing the dataloader
- efficient video decoding on the fly
- training/testing the model
- training procedure using precomputed masks
You can download the precomputed masks using the links at the top of the page. The resolution is of size 100x100 and we threshold the predictions with a minimum confidence of 0.5.
- pytorch 0.4.0
- numpy
- lintel - make sure that you have already installed this library (important for decoding videos on the fly)
If you find this paper or our implementation useful for your research or if you use the precomputed masks, please cite our paper.
@InProceedings{Baradel_2018_ECCV,
author = {Baradel, Fabien and Neverova, Natalia and Wolf, Christian and Mille, Julien and Mori, Greg},
title = {Object Level Visual Reasoning in Videos},
booktitle = {ECCV},
year = {2018}
}
This work was funded by grant Deepvision (ANR-15- CE23-0029, STPGP-479356-15), a joint French/Canadian call by ANR & NSERC.
MIT License