Datasets For Scene Analysis and Reconstruction

Datasets For Scene Analysis and Reconstruction

Object Detection and Tracking Datasets

DatasetYearLinkDataset SizeType of Access
MIT pedestrian dataset2000MIT pedestrian22MBPublic
Caltech Pedestrian Dataset2009Caltech11,9 GBPublic
INRIA Pedestrian Dataset2005INRIA1,1 GBPublic
PASCAL Visual Object Classes 20072007Pascal VOC 2007880MBPublic
PASCAL Visual Object Classes 20122012Pascal VOC 20122GB Train/Validation DataRegister Required for Test Data
ImageNet Large Scale Visual Recognition Challenge (ILSVRC)2015ILSVRC167.62 GBAvailable on Kaggle. Additional data requires registration.
Microsoft COCO (MS-COCO)2014MS-COCO62GB all labeled imagesPublic
Open Images2017Open Images561GBPublic

Visual Relationships and Scene Graph Generation

DatasetYearLinkDataset SizeType of Access
Real-World Scene Graphs Dataset2015Real-World SG2GBPublic
Visual Relationship Dataset (VRD)2016VRD1.86GBPublic
Visual Genome Dataset (VGD)2017Visual Genome15.7GBPublic
HCVRD Dataset2018HCVRD15.7GBPublic

Action Recognition

DatasetYearLinkDataset SizeType of Access
Sports1M2014Sports1M1.13m youtube videosPublic
YouTube8M2016YouTube8M31GB for video features, 1.53TB for frame featuresPublic
Charades2016Charades55 GBPrivate?
Kinetics2017Kinetics650k youtube videosPublic
AVA Actions Datase2018AVA430 15-minute youtube videosPublic
Moments in Time2019MiT1m 3s videosRequest Required
HACS2019HACS50k youtube videosPublic
Holistic Video Understanding2020HVU577k youtube videosPublic for train set, Request for test and missing videos
BABEL2021BABEL43 hours of MOCAP sequences from the AMASS datasetRegistration Required

Human Attributes Recognition

DatasetYearLinkDataset SizeType of Access
Parse 27K2015Parse 27K3.7GBPublic
RAP-2.02018RAP 2.084928 images of pedestriansRegistration Required
Human Attributes (HAT)2011HAT9344 images of pedestriansContact Required
Berkeley-Attributes of People (BAP)2011BAP671MBPublic
PA-100K2017PA 100k430MBPublic

Pose Estimation

DatasetYearLinkDataset SizeType of Access
Leeds Sports Pose (LSP)2010LSP163MBPublic
MPII Dataset2014MPII12.9GBPublic
Joint-annotated HMDB (J-HMDB)2013J-HMDB928 clips with 21 action categoriesRegistration Required
Human-in-Events (HiEve)2020HiEVE1M+ poses and 56k+ action labelsPublic
Human3.6M2013Human3.6M3.6 million human poses and corresponding imagesAccount creation required
MoVi2020MoVi1056 files with MoCap and video dataAccess request required
AMASS2019AMASS40 hours of motion dataRegistration required
3DPW20183DPW60 video sequences with 2D and 3D posesPublic
CMU Panoptic2015CMU PanopticOffline
Joint Track Auto (JTA)2018JTA12.1GBPublic

Gait Estimation

DatasetYearLinkDataset SizeType of Access
CASIA2006CASIA10GBRegistration Required
GREW2021GREW233 857 sequencesRegistration Required

Scene Reconstruction

DatasetYearLinkDataset SizeType of Access
Sun RGB-D2015SUN RGB-D6.4GBPublic
Scan Net2017Scan-net1500 scansAccess request required
SURREAL2017Surreal86GBAccept the license terms
Scan2CAD2019Scan2CAD1506 scansRegistration Required
Kitti-3602021Kitti-360 320k images and 100k laser scans in a driving distance of 73.7kmRegistration Required
RELLIS-3D2020RELLIS-3D13,556 LiDAR scans and 6,235 imagesPublic
Hypersim2021Hypersim1.9TBPublic, but the ground truth triangle meshes for each scene needs to be purchased