32 datasets found

Person re-identification (re-id) is the problem of matching people across non-overlapping camera views distributed in open spaces at different locations, typically achieved by...

The PRID2011 dataset extracts images from video recorded by static surveillance cameras with two camera views.

PETA dataset is a large-scale person re-identification dataset that contains binary and multi-class attribute annotations.

Person Re-ID has two key steps: obtaining a feature embedding and performing matching under some distance metric.

A runner re-identification system for a single-view video in the open-world setting.

The Duke dataset is a large-scale person re-identification dataset, containing 702 identities with 16522 images.

Person re-identification (re-ID) is a fundamental task in distributed multi-camera surveillance, aiming to match people appearing in different non-overlapping camera views.

The dataset contains 171,352 images of 1,082 pedestrian identities.

The dataset contains 19,060 images of 512 identities captured from 4 cameras.

Person re-identification (re-ID) under various occlusions has been a long-standing challenge as person images with different types of occlusions often suffer from misalign-ment...

Person Re-Id benchmark datasets for evaluation of person re-identification methods

Person search unifies person detection and person re-identification (Re-ID) to locate query persons from the panoramic gallery images. One major challenge comes from the...

CUHK-SYSU is a large-scale dataset designed for person search, which contains 18,184 scene images captured from street nap and movie screenshot.

The dataset used for tracking multiple people across multiple cameras is a large-scale multi-camera tracking dataset, DukeMTMC.

A synthetic dataset for pedestrian tracking and Re-ID with highly accurate and rich labels of ids, 2D bounding box coordinates, and semantic attributes.

Recognizing person in UAV images faces two main challenges: significant size difference and uncertain rotation variations.

A large-scale, long-term, person re-ID dataset, called Celeb-reID, was built in 2019.

The Market-1501 dataset is a large benchmark for person re-identification. It contains 12,000 images with 1,500 images per person, annotated with bounding boxes and masks for...

Person re-identification by deep joint learning of multi-loss classification.

SpindleNet for learning person features of different body regions with additional human pose information.