Multi-Spectral Stereo ($MS^2$) Outdoor Driving Dataset

This is the official github page of the $MS^2$ dataset described in the following paper.

This page provides a dataloader and simple python code for $MS^2$ dataset.

If you want to download the dataset and see the details, please visit the dataset page.

Deep Depth Estimation from Thermal Image

Ukcheol Shin, Jinsun Park, In So Kweon

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023

[Paper] [Dataset page]

Updates

• 2023.03.30: Open Github page.
• 2023.05.30: Release $MS^2$ dataset, dataloader, and demo code.

$MS^2$ Dataset Specification

MS2 dataset provides:

• (Synchronized) Stereo RGB images / Stereo NIR images / Stereo thermal images
• (Synchronized) Stereo LiDAR scans / GPS/IMU navigation data
• Projected depth map (in RGB, NIR, thermal image planes)
• Odometry data (in RGB, NIR, thermal cameras, and LiDAR coordinates)

Usage

1. Download the datasets and place them in 'MS2dataset' folder in the following structure:

MS2dataset
├── sync_data
│   ├── <Sequence Name1>
│   ├── <Sequence Name2>
│   ├── ...
│   └── <Sequence NameN>
├── proj_depth
│   ├── <Sequence Name1>
│   ├── <Sequence Name2>
│   ├── ...
│   └── <Sequence NameN>
└── odom
    ├── <Sequence Name1>
    ├── <Sequence Name2>
    ├── ...
    └── <Sequence NameN>

2. We provide a simple python code (demo.py) along with a dataloader to take a look at the provided dataset. To run the code, you need any version of Pytorch library.

   python demo.py --seq_name <Sequence Name> --modality rgb --data_format MonoDepth
   python demo.py --seq_name <Sequence Name> --modality nir --data_format StereoMatch
   python demo.py --seq_name <Sequence Name> --modality thr --data_format MultiViewImg