Dataset Library

A curated catalog of vision datasets across every domain — with size, task, and license at a glance.

Classification
LAION-5B
Image–text pretraining

Web-scale image-text dataset behind CLIP-style and diffusion models.

5.85B image–text pairs
CC BY 4.0 (metadata)· 2022
Medical
BraTS
Brain tumor segmentation

Multi-institutional MRI for 3D brain tumor segmentation challenges.

Multi-modal MRI volumes
Research use· 2020
Autonomous
nuScenes
3D detection & tracking

Full sensor-suite (camera, LiDAR, radar) driving dataset for 3D perception.

1000 scenes · 1.4M images
CC BY-NC-SA· 2019
Medical
ISIC (Skin Lesion)
Lesion classification & segmentation

Dermoscopy images for melanoma detection — a standard medical imaging benchmark.

~25K dermoscopic images
CC-0 / varies· 2019
3D & Synthetic
MVTec AD
Industrial anomaly detection

Defect/anomaly detection benchmark for industrial quality inspection.

5K images · 15 categories
CC BY-NC-SA· 2019
Satellite
DOTA
Aerial oriented detection

Large aerial images with oriented bounding boxes for remote sensing.

2.8K images · 18 classes
Research use· 2018
Detection
Open Images
Detection & relationships

Massive, richly annotated detection dataset with hierarchical labels.

9M images · 600 classes
CC BY 4.0· 2017
Segmentation
ADE20K
Scene parsing

Dense semantic segmentation across diverse scenes and stuff/thing classes.

25K images · 150 classes
Research use· 2017
OCR
ICDAR / SynthText
Text detection & recognition

Scene-text detection and recognition benchmarks, real and synthetic.

800K synthetic + real
Research use· 2017
Video
Kinetics-400/700
Action recognition

Large-scale human action video dataset for spatiotemporal models.

650K clips · 700 classes
CC BY 4.0· 2017
Autonomous
Cityscapes
Urban scene segmentation

High-quality pixel annotations of street scenes for driving perception.

25K images · 30 classes
Research use· 2016
3D & Synthetic
ShapeNet
3D shape understanding

Richly annotated 3D CAD models for reconstruction and 3D vision.

51K 3D models · 55 categories
Research use· 2015
Detection
MS COCO
Detection, segmentation, captioning

The de-facto benchmark for object detection and instance segmentation.

330K images · 80 classes
CC BY 4.0· 2014
Detection
Pascal VOC
Detection & segmentation

A classic, compact detection/segmentation benchmark — great for learning.

~11K images · 20 classes
Research use· 2012
Autonomous
KITTI
Detection, depth, odometry

Foundational autonomous-driving benchmark with LiDAR + stereo + boxes.

~15K frames
CC BY-NC-SA· 2012
Classification
ImageNet-1K
Image classification

The benchmark that launched deep learning for vision; standard pretraining dataset.

1.28M images · 1000 classes
Research use· 2009