Active-Passive SimStereo

Dataset content

The dataset contains 515 image pairs, pre-split into a test set of 103 image pairs and a train set of 412 image pairs.

The images have a standard resolution of 640x480 pixels.

The provided ground truth is precise up to floating point precision.

Raw images and ground truth disparities

The raw images in linear color space, as well as the ground truth disparities, are provided as exr images with 32bits floating point precision.

To read the different layers, any exr library should work fine. We are using the python module provided by the exr-tools project by french-paragon:
https://github.com/french-paragon/exr-tools

You can also inspect the content of the images with the embeded image viewer in blender or any software capable of reading exr images.

• The left and right images are contained in the Left, Respectively Right, Layers.
• The Color, Nir and Disparity Images are contained in the Color, Nir and Disp passes.
• The Color passe is made up of the standard R, G and B channels
• The Nir pass is made up of a single A channel
• The Disp pass is made up of a single D channel

Additional Disparity formats

To make the ground truth disparities easier to use in a variety of situations, we do provide them in npy (python numpy) and pfm (portable float map) formats as well.

Color managed images

We also exported color managed perceptual images (sRGB instead of RGB Linear color space) in standard jpg format if you prefer to use these.

The Images are in the rgbColormanaged and nirColormanaged folder. Each image pair comprises two images with the same name, but the left or right suffix. You can open them with any software or library able to read .jpg images

Dataset paper

If you use our dataset, please cite our paper:

@inproceedings{
jospin2022activepassive,
title={Active-Passive SimStereo - Benchmarking the Cross-Generalization Capabilities of Deep Learning-based Stereo Methods},
author={Laurent Valentin Jospin and Allen Antony and Lian Xu and Hamid Laga and Farid Boussaid and Mohammed Bennamoun},
booktitle={Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
year={2022},
url={https://openreview.net/forum?id=ml1NjI-ujzf}
}

[EDIT 04-26-2022] Simulation toolkit

We made the simulation toolkit we used to produce the dataset available so that you can create your own images if you need to. See the simulation.zip file.

[EDIT 06-15-2022] Additional formats for the disparities

We added more image formats for the ground truth disparities to make them easier to use in a variety of scenarios.

[EDIT 08-26-2022] Additional data packs

We released an additional data pack containing 5 sequences of stereo images. More data packs will be released in the future. These data packs are optimized for quantity rather than diversity and as such are based on large sequences of images rather than punctual views of scenes.