Awesome

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

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>

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