Awesome Multispectral Datasets

A curated list of publicly available multispectral datasets, tools, and resources for computer vision and remote sensing.

Contents

• Dataset catalog (table)
• Categorized by Topic
• Categorized by Task
• Categorized by Electromagnetic Spectrum
• Contributing

Dataset catalog (table)

Short name	Affiliation	Year	Modality	Collection platform	Data Type	Dataset size	Environment	Task(s)	Publication venue	link
KAIST Multispectral Pedestrian	KAIST	2015	Thermal	Vehicle	Image	95,328 frames; 103,128 objects	Urban; day/night	Pedestrian detection	CVPR 2015	Project Page
FLIR ADAS Thermal	Teledyne FLIR	2018	Thermal	Vehicle	Image & Video	10K+ images (v1); videos in v2	Urban/suburban; day/night	Object detection	—	Project Page
MFNet (RGB–Thermal Segmentation)	ETRI / Univ. of Ulsan	2018	Thermal	Vehicle	Image	1,569 labeled frames	Urban; day/night	Semantic segmentation	—	Github
MSRS (Multispectral Road Scenes)	—	2020		Vehicle	Image	3K+ RGB–Thermal pairs	Urban roads; day/night	Semantic segmentation	—	Github
Seeing Through Fog	Princeton	2020		Vehicle	Image	10K+ samples	Adverse weather	Detection; depth; segmentation	IJRR 2020	Project Page
DENSE (All‑weather multimodal)	TUM / DENSE EU	2020		Vehicle	Multi	Multi‑sensor sequences	Adverse weather	Detection; tracking; depth	IJRR 2020	Project Page
MS2 (RGB/NIR/Thermal)	KAIST	2023	NIR, Thermal	Vehicle	Image	184K data pairs	City; campus; suburban	Stereo; depth; fusion	CVPR 2023	Project Page
LLVIP (Low‑Light Visible‑Infrared Paired)	—	2021	Thermal	Surveillance	Image	15,488 pairs	Low‑light; indoor/outdoor	Detection; enhancement	ICCV 2021	Dataset
M3FD	—	2022		Vehicle	Image	4,200 images	Overcast; urban	Object detection	CVPR 2022	Summary
CATS (Color‑Thermal Stereo)	—	2017	Thermal	Handheld/rig		1,400 pairs	Indoor/outdoor	Stereo; calibration	CVPR 2017	Project Page
TNO Multiband	TNO	2014		Static		60 pairs	Military; surveillance	Fusion; detection	Data in Brief 2017	Dataset
RoadScene (FusionDN)	—	2020		Vehicle		221 pairs	Driving; road	Image fusion	AAAI 2020	Github
MVSeg	—	2023	Thermal	Vehicle	Video	53K frames (738 videos)	Driving; urban	Semantic segmentation	CVPR 2023	Project Page
RASMD	Yonsei	2024	SWIR	Vehicle	Image	100,000 pair of images	Urban, Suburban	Object Detection, Super Resolution, I2I translation	Arxiv	Project Page
MCubeS	Kyoto	2022	Polarization, NIR	Custom	Image	500	Outdoor	Semantic segmentation	CVPR 2022	Github
SemanticRT	—	2023		Vehicle	Image	11,371 images	Urban	Semantic segmentation	ACM MM 2023	Github

Notes:

• Counts are approximate; see the linked pages for authoritative details.
• Prefer citing the official site if available; otherwise link to a stable summary.

Categorized by Topic

• Autonomous driving
  • KAIST Multispectral Pedestrian — RGB+Thermal; pedestrian detection. Link: https://sites.google.com/view/multispectral/object/pedestrian-2015
  • FLIR ADAS Thermal — RGB+Thermal; detection in day/night. Link: https://www.flir.com/oem/adas/adas-dataset/
  • MFNet — RGB+Thermal; semantic segmentation. Link: https://github.com/haqishen/MFNet-pytorch#dataset
  • MSRS — RGB+Thermal; semantic segmentation. Link: https://github.com/Linfeng-Tang/MSRS
  • Seeing Through Fog — multimodal incl. thermal; adverse weather. Link: https://light.princeton.edu/seeing-through-fog/
  • DENSE — multimodal incl. thermal; all‑weather. Link: https://www.dense-dataset.com/
  • LLVIP — visible+infrared low‑light pairs; detection/enhancement. Link: https://github.com/bupt-ai-cz/LLVIP
  • M3FD — visible+infrared for detection. Link: https://github.com/JinyuanLiu-CV/TarDAL
  • MVSeg — multispectral video segmentation. Link: https://jiwei0921.github.io/Multispectral-Video-Semantic-Segmentation/
  • SemanticRT — multispectral segmentation. Link: https://github.com/jiwei0921/SemanticRT
  • MS2 — RGB/NIR/Thermal; stereo/depth. Link: https://sites.google.com/view/multi-spectral-stereo-dataset

Categorized by Task

• Detection

  • RASMD — RGB and SWIR autonomous driving dataset. Link: https://yonsei-stl.github.io/RASMD/
  • KAIST — Multispectral Pedestrian — RGB+Thermal; pedestrian detection. Link: https://sites.google.com/view/multispectral/object/pedestrian-2015
  • FLIR — ADAS Thermal — RGB+Thermal; detection. Link: https://www.flir.com/oem/adas/adas-dataset/
  • M3FD — RGB+Thermal; object detection. Link: https://github.com/JinyuanLiu-CV/TarDAL
  • LLVIP — visible+infrared; low‑light detection. Link: https://github.com/bupt-ai-cz/LLVIP
  • Seeing Through Fog — multimodal incl. thermal; object detection. Link: https://light.princeton.edu/seeing-through-fog/
  • DENSE — multimodal incl. thermal; object detection/tracking. Link: https://www.dense-dataset.com/

• Semantic segmentation

  • MFNet — RGB+Thermal segmentation. Link: https://github.com/haqishen/MFNet-pytorch#dataset
  • MSRS — RGB+Thermal segmentation. Link: https://github.com/Linfeng-Tang/MSRS
  • MVSeg — multispectral video segmentation. Link: https://jiwei0921.github.io/Multispectral-Video-Semantic-Segmentation/
  • SemanticRT — multispectral segmentation. Link: https://github.com/jiwei0921/SemanticRT
  • DeepScene — RGB+NIR; forest segmentation. Link: https://arxiv.org/abs/1703.01918

• Stereo/depth

  • MS2 — RGB/NIR/Thermal; stereo/depth. Link: https://sites.google.com/view/multi-spectral-stereo-dataset
  • CATS — Color‑Thermal stereo for calibration/matching. Link: https://openaccess.thecvf.com/content_cvpr_2017/html/Bilodeau_CATS_A_Color_CVPR_2017_paper.html
  • Seeing Through Fog — multimodal; depth estimation. Link: https://light.princeton.edu/seeing-through-fog/

• Fusion/registration

  • TNO Multiband — visible+thermal image fusion. Link: https://figshare.com/articles/dataset/TNO_Image_Fusion_Dataset/1008029
  • RoadScene (FusionDN) — visible+thermal image fusion. Link: https://github.com/hli1221/FusionDN
  • CATS — color+thermal stereo pairs. Link: https://openaccess.thecvf.com/content_cvpr_2017/html/Bilodeau_CATS_A_Color_CVPR_2017_paper.html

• Remote sensing (satellite)

  • Sentinel‑2 MSI — Multispectral optical; global coverage. Link: https://dataspace.copernicus.eu/
  • Landsat 8/9 OLI/TIRS — Multispectral optical/thermal. Links: https://earthexplorer.usgs.gov/, https://registry.opendata.aws/landsat-c2/
  • MODIS — Multispectral products; long time series. Link: https://ladsweb.modaps.eosdis.nasa.gov/

Categorized by Electromagnetic Spectrum

• Near‑Infrared (NIR)

  • MS2 — RGB/NIR/Thermal; stereo/depth. Link: https://sites.google.com/view/multi-spectral-stereo-dataset

• Short‑Wave Infrared (SWIR)

  • RASMD — RGB+SWIR driving dataset. Link: https://arxiv.org/abs/2504.07603

• Thermal Infrared (LWIR)

  • KAIST Multispectral Pedestrian — RGB+Thermal. Link: https://sites.google.com/view/multispectral/object/pedestrian-2015
  • FLIR ADAS Thermal — RGB+Thermal. Link: https://www.flir.com/oem/adas/adas-dataset/
  • MFNet — RGB+Thermal segmentation. Link: https://github.com/haqishen/MFNet-pytorch#dataset
  • Seeing Through Fog — includes thermal modality. Link: https://light.princeton.edu/seeing-through-fog/
  • DENSE — includes thermal modality. Link: https://www.dense-dataset.com/

Contributing

PRs welcome! Please follow the Awesome List style:

• Keep sections alphabetized where possible.
• For each entry, include: name — one‑sentence description — key properties (bands, resolution) — link(s).
• Prefer official or long‑term stable links (program portals, data registries).
• Ensure the dataset is publicly accessible and note any access requirements.

Suggested entry format:

- Dataset Name — short description; key properties (e.g., bands, resolution). Link: https://... (and/or additional link)