Person Re-Identification

SenseReID is a person re-identification dataset for evaluating ReID models. It is captured from real surveillance cameras and the person bounding boxes are obtained from state-of-the-art detection algorithm. The dataset contains 1,717 identities in total.

The ClonedPerson dataset is a large-scale synthetic person re-identification dataset introduced in the paper "Cloning Outfits from Real-World Images to 3D Characters for Generalizable Person Re-Identification" in CVPR 2022. It is generated by MakeHuman and Unity3D. Characters in this dataset use an automatic approach to directly clone the whole outfits from real-world person images to virtual 3D characters, such that any virtual person thus created will appear very similar to its real-world counterpart. The dataset contains 887,766 synthesized person images of 5,621 identities.

iQIYI-VID dataset, which comprises video clips from iQIYI variety shows, films, and television dramas. The whole dataset contains 500,000 videos clips of 5,000 celebrities. The length of each video is 1~30 seconds. Source: iQIYI-VID

Person re-ID matches persons across multiple non-overlapping cameras. Despite the increasing deployment of airborne platforms in surveillance, current existing person re-ID benchmarks' focus is on ground-ground matching and very limited efforts on aerial-aerial matching. We propose a new benchmark dataset - AG-ReID, which performs person re-ID matching in a new setting: across aerial and ground cameras. Our dataset contains 21,983 images of 388 identities and 15 soft attributes for each identity. The data was collected by a UAV flying at altitudes between 15 to 45 meters and a ground-based CCTV camera on a university campus. Our dataset presents a novel elevated-viewpoint challenge for person re-ID due to the significant difference in person appearance across these cameras.

Large, multimodal biometric dataset: It contains still images and videos of over 1,000 people captured at various ranges (up to 1,000 meters) and elevations (up to 400 meters) using a diverse set of cameras (commercial, military-grade, specialized). Focus on challenging scenarios: Designed to address limitations of existing datasets, it includes data with extreme poses, low resolutions, and atmospheric distortions to push the boundaries of biometric recognition algorithms. Rich annotations and diversity: Each data point is accompanied by detailed metadata (sensor details, weather conditions) and manual annotations (bounding boxes, yaw/pitch angles) for face and whole-body regions. The dataset also ensures demographic diversity for robust algorithm development.

The Market1501-Attributes dataset is built from the Market1501 dataset. Market1501 Attribute is an augmentation of this dataset with 28 hand annotated attributes, such as gender, age, sleeve length, flags for items carried as well as upper clothes colors and lower clothes colors. Source: Color inference from semantic labeling for person search in videos Image Source: https://github.com/vana77/Market-1501_Attribute

MSMT17-C is an evaluation set that consists of algorithmically generated corruptions applied to the MSMT17 test-set. These corruptions consist of Noise: Gaussian, shot, impulse, and speckle; Blur: defocus, frosted glass, motion, zoom, and Gaussian; Weather: snow, frost, fog, brightness, spatter, and rain; Digital: contrast, elastic, pixel, JPEG compression, and saturate. Each corruption has five severity levels, resulting in 100 distinct corruptions.

Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions. Source: The P-DESTRE: A Fully Annotated Dataset for Pedestrian Detection, Tracking, Re-Identification and Search from Aerial Devices

This dataset contains 114 individuals including 1824 images captured from two disjoint camera views. For each person, eight images are captured from eight different orientations under one camera view and are normalized to 128x48 pixels. This dataset is also split into two parts randomly. One contains 57 individuals for training, and the other contains 57 individuals for testing. Source: https://github.com/charliememory/PKU-Reid-Dataset Image Source: https://arxiv.org/pdf/1605.02464.pdf

The images in DukeMTMC-attribute dataset comes from Duke University. There are 1812 identities and 34183 annotated bounding boxes in the DukeMTMC-attribute dataset. This dataset contains 702 identities for training and 1110 identities for testing, corresponding to 16522 and 17661 images respectively. The attributes are annotated in the identity level, every image in this dataset is annotated with 23 attributes. NOTE: This dataset has been retracted. Source: Pedestrian Attribute Recognition: A Survey Image Source: http://irip.buaa.edu.cn/mars_duke_attributes/index.html

The eSports Sensors dataset contains sensor data collected from 10 players in 22 matches in League of Legends. The sensor data collected includes: * Hand/head/chair movements. * Heart rate. * Muscle activity. * Gaze movement on the monitor. * Galvanic skin response(GSR). * Electroencephalography (EEG). * Mouse and keyboard activity. * Facial skin temperature. * Environmental data. The data were collected for one team of 5 people simultaneously. In-game logs and meta information for each match are also provided for each match. Source: https://github.com/smerdov/eSports_Sensors_Dataset

Multi-view Extended Videos with Identities dataset (MEVID) is a dataset for large-scale, video person re-identification (ReID) in the wild. It spans an extensive indoor and outdoor environment across nine unique dates in a 73-day window, various camera viewpoints, and entity clothing changes. Specifically, it contains labels of the identities of 158 unique people wearing 598 outfits taken from 8, 092 tracklets, average length of about 590 frames, seen in 33 camera views from the very large-scale MEVA person activities dataset. Source: MEVID: Multi-view Extended Videos with Identities for Video Person Re-Identification Image Source: https://arxiv.org/pdf/2211.04656v1.pdf

RegDB-C is an evaluation set that consists of algorithmically generated corruptions applied to the RegDB test-set (color images). These corruptions consist of Noise: Gaussian, shot, impulse, and speckle; Blur: defocus, frosted glass, motion, zoom, and Gaussian; Weather: snow, frost, fog, brightness, spatter, and rain; Digital: contrast, elastic, pixel, JPEG compression, and saturate. Each corruption has five severity levels, resulting in 100 distinct corruptions.

SYSU-MM01-C is an evaluation set that consists of algorithmically generated corruptions applied to the SYSU-MM01 test-set. These corruptions consist of Noise: Gaussian, shot, impulse, and speckle; Blur: defocus, frosted glass, motion, zoom, and Gaussian; Weather: snow, frost, fog, brightness, spatter, and rain; Digital: contrast, elastic, pixel, JPEG compression, and saturate. Each corruption has five severity levels, resulting in 100 distinct corruptions.

The OpeReid dataset is a person re-identification dataset that consists of 7,413 images of 200 persons. Source: Scalable Metric Learning via Weighted Approximate Rank Component Analysis

Aerial-ground person re-identification (Re-ID) presents unique challenges in computer vision, stemming from the distinct differences in viewpoints, poses, and resolutions between high-altitude aerial and ground-based cameras. Existing research predominantly focuses on ground-to-ground matching, with aerial matching less explored due to a dearth of comprehensive datasets. To address this, we introduce AG-ReID.v2, a dataset specifically designed for person Re-ID in mixed aerial and ground scenarios. This dataset comprises 100,502 images of 1,615 unique individuals, each annotated with matching IDs and 15 soft attribute labels. Data were collected from diverse perspectives using a UAV, stationary CCTV, and smart glasses-integrated camera, providing a rich variety of intra-identity variations.

This dataset contains 1203 individuals captured from two disjoint camera views. To each person, one to twelve images are captured from one to six different orientations under one camera view and are normalized to 128x64 pixels. This dataset is constructed based on the Market-1501 benchmark data and the orientation label for each image has been manually annotated. Source: https://github.com/charliememory/Market1203-Reid-Dataset

The 42Street dataset is based on a theater play as an example of such an application. The dataset is created using a public recording of the 42Street theatre play [42street]. The play is 1.5 hours long and was split into 5 equally long parts of 20 minutes each, with various clothes changes between the different parts. Source: ReFace: Improving Clothes-Changing Re-Identification With Face Features Image Source: https://arxiv.org/pdf/2211.13807v1.pdf

The largest aerial-ground video dataset with 6,632 identities, 32,321 tracklets, and 9.6 million frames Multiple platforms combining aerial drones (at various altitudes: 15m, 30m, 80m, 120m), CCTV cameras, and wearable cameras Rich annotations including clothing changes and 15 soft-biometric attributes Real-world challenges including extreme viewpoint variations, scale differences, and occlusions Diverse environments with outdoor campus scenes captured over 20 days

Correlated Corrupted Dataset is an evaluation set that consists of realistic visible-infrared (V-I) corruptions allowing for models' corruption robustness evaluation. Initially proposed for multimodal person re-identification, our dataset can also be used for the evaluation of V-I cross-modal approaches. Corruptions of the visible modality are the twenty corruptions proposed by Chen & al. in the "Benchmarks for Corruption Invariant Person Re-identification" paper. Corruptions of the infrared modalities have been proposed in our paper, introducing 19 corruptions that respect the infrared modality encoding. In practice, for co-located visible-infrared cameras, weather-related corruptions should, for example, affect each camera. Also, blur-related corruption would likely occur in both visible and infrared cameras. This dataset tackles this aspect by considering the eventual correlations that may occur from one modality camera to another.

This dataset contains 971 identities from two disjoint camera views. Each identity has two samples per camera view. It is used for Person Re-identification. Paper: Li W., Zhao R., Wang X. (2013) Human Reidentification with Transferred Metric Learning. In: Lee K.M., Matsushita Y., Rehg J.M., Hu Z. (eds) Computer Vision – ACCV 2012. ACCV 2012. Lecture Notes in Computer Science, vol 7724. Springer, Berlin, Heidelberg

Collected data from two distinct experiments in immersive, interactive VR where participants performed dynamic tasks as their eye, head, and hand movements were recorded. In the second experiment, a range of real-time privacy mechanisms are applied to eye gaze in real-time.

The growing importance of person re-identification in computer vision has highlighted the need for more extensive and diverse datasets. In response, we introduce the ENTIRe-ID dataset, an extensive collection comprising over 4.45 million images from 37 different cameras in varied environments. This dataset is uniquely designed to tackle the challenges of domain variability and model generalization, areas where existing datasets for person re-identification have fallen short. The ENTIRe-ID dataset stands out for its coverage of a wide array of real-world scenarios, encompassing various lighting conditions, angles of view, and diverse human activities. This design ensures a realistic and robust training platform for ReID models.

The IUST_PersonReID dataset was developed to address limitations in existing person re-identification datasets by including cultural and environmental contexts unique to Islamic countries, especially Iran and Iraq. Unlike common datasets, which don’t reflect the clothing styles common in these regions—such as hijabs and other coverings—the IUST_PersonReID dataset represents this diversity, helping to reduce demographic bias and improve model accuracy. Collected from various real-world settings under different lighting, camera angles, indoor & outdoor, and weather conditions, this dataset provides extensive, overlapping views across multiple cameras. By capturing these unique conditions, IUST_PersonReID offers a valuable resource for developing re-ID models that perform more reliably across diverse environments and populations.

MARS dataset processed with our re-Detect and Link (DL) module. More information: https://github.com/jackie840129/CF-AAN

MVB (Multi View Baggage) is a dataset for baggage ReID task which has some essential differences from person ReID. The features of MVB are three-fold. First, MVB is the first publicly released large-scale dataset that contains 4519 baggage identities and 22660 annotated baggage images as well as its surface material labels. Second, all baggage images are captured by specially-designed multi-view camera system to handle pose variation and occlusion, in order to obtain the 3D information of baggage surface as complete as possible. Third, MVB has remarkable inter-class similarity and intra-class dissimilarity, considering the fact that baggage might have very similar appearance while the data is collected in two real airport environments, where imaging factors varies significantly from each other. Source: MVB: A Large-Scale Dataset for Baggage Re-Identification and Merged Siamese Networks

We introduce Occluded PoseTrack-ReID (or simply Occ-PTrack), a new ReID dataset we built out of the annotation available with PoseTrack21, a popular video benchmark for multi-person pose tracking, that features keypoints and cross-video identity annotations. Unlike previous ReID datasets focused on street surveillance, Occ-PTrack consists of images from everyday life videos, primarily from sports activities. Occ-PTrack is divided into a train/test that includes 1000/1411 identities with 17.898/13.412 images from 474/170 videos, which is roughly equivalent in terms of scale to other popular ReID datasets (e.g. Market-1501, Occluded-Duke, ...). To assess the ReID model’s performance in multi-person occlusion scenarios, we select the most cluttered images of each identity in the test set as query samples, and the remaining test images as gallery samples. Cluttered images corresponds to multi-persons occlusions scenarios where either the front (occluding) or back (occluded) person is the ReID target, to evaluate the model ability to re-identify individuals in both scenarios. We provide further details about our proposed dataset in the supp. materials. Occ-PTrack is challenging as persons within the same video exhibit a high degree of visual resemblance since they often wear similar sports kits.

The TVPReid dataset contains 6559 pedestrian videos, each of which is annotated with two text descriptions, for a total of 13118 descriptions. The sentence descriptions are in a natural language style and contain rich details about the pedestrian's appearance, actions, and environmental elements that the pedestrian interacts with. The average sentence length of the TVPReid dataset is 30 words, and the longest sentence contains 83 words.

The ULI-RI dataset is generated using the Unreal Engine 4 to simulate various outdoor environments with 115 high-quality 3D human models. For each person identity, we controlled and quantitatively labeled the illumination intensity, view point (model z-rotation angle), and background to create 512 images. There are total 115 x 512 = 58880 images in the ULI-RI dataset.

Uncorrelated Corrupted Dataset is an evaluation set that consists of realistic visible-infrared (V-I) corruptions allowing for models' corruption robustness evaluation. Initially proposed for multimodal person re-identification, our dataset can also be used for the evaluation of V-I cross-modal approaches. Corruptions of the visible modality are the twenty corruptions proposed by Chen & al. in the "Benchmarks for Corruption Invariant Person Re-identification" paper. Corruptions of the infrared modalities have been proposed in our paper, introducing 19 corruptions that respect the infrared modality encoding. In practice, the corruptions are applied randomly and independently to the visible and the infrared cameras, making it more suited to a not co-located camera setting.

The X-MARS dataset proposes new splits for the MARS dataset, to allow for cross-evaluation with the Market-1501 dataset without training and test overlap between the two datasets. Source: https://github.com/andreas-eberle/x-mars