Datasets

A large scale natively low-resolution face recognition benchmark. It consists of 169,403 native low-resolution face images (average 20×16 pixels) from 5,139 distinct identities. All the face images in TinyFace were collected from public web data across a large variety of imaging scenarios, captured under uncontrolled viewing conditions in pose, illumination, occlusion and background.

A challenging vehicle re-idendification benchmark, characterised by vehicle images subject to more realistic and unconstrained variations in resolution (scale), motion blur, illumination, occlusion, and viewpoint, and background clutter. It contains 60,434 images of 5,626 vehicle identities captured by 60 different cameras at heterogeneous road traffic scenes in both day-time and night-time.

A large scale Open Logo Detection Challenge: only a small proportion of logo classes provide fine-grained object bounding box labelling whilst the remaining classes have no labelled training data. This simulates a realistic logo detection scenario where new logo classes arrive progressively and require to be detected with little or none budget for exhaustively labelling fine-grained training data for every new class. This benchmark contains 27,083 images from 352 unique logo classes.

A large scale Surveillance Face Recognition Challenge, where low-resolution face images are not synthesised by artificial down-sampling of native high-resolution images. This benchmark contains 463,507 face images of 15,573 distinct identities captured in real-world uncooperative surveillance scenes over wide space and time.

A large scale weakly and noisely labelled Logo Detection dataset consisting of (1) over 2 million web images and (2) 6,000+ test images with manually labelled logo bounding boxes.

A Logo Detection dataset containing 10 most popular brand logos of shoes, clothing and accessories.

An image sequence based person re-identification dataset captured in a crowded public space.

A busy outdoor dataset for research on visual surveillance.

An indoor dataset collected from a university campus for physical event understanding of long video streams.