An overview of the GPR+ dataset.
We provide coarse-to-fine attribute annotations both temporally and semantically, including identity, viewpoint, weather and illumination. In order to make the attributes more independent and complementary to each, we redefined the attribute distribution in this upgraded version GPR+.
Paper
Analysis
GitHub Repo
Abstract
Person re-identification (re-ID) plays an important role in
applications such as public security and video surveillance.
Recently, learning from synthetic data, which benefits from
the popularity of synthetic data engine, has achieved remarkable
performance. However, existing methods only focus on
achieving high performance on re-ID task with large-scale
dataset while ignoring that how to perform efficient training
from millions of synthetic data. To facilitate development
in this field, we reviewed the previously developed synthetic
dataset GPR and built an improved one (GPR+) with larger
number of identities and distinguished attributes. This new dataset
will help us have a deeper understanding of the fundamental
problems in person re-ID. Our research also provides useful
insights for dataset building and future practical usage. Note that
GPR+ will be released with a open source license to enable further developments
in person re-identification in the next few months.
We present a detailed comparison between existing synthetic datasets (e.g. SOMAset, SyRI, PersonX and GPR) and our upgraded version GPR+. In particular, "#view" indicates whether dataset has viewpoint attribute labels, simultaneously for "#weather" and "#illumination".
Identity
We present the upgraded GPR+ dataset which consists of 808 identities and 475,104 bounding boxes, with resolution of 200 × 470, including
more high-quality attribute annotations.
Viewpoint
We present GPR+ consisting of 12 different types of viewpoints, every 30° from 0° to 330°.
Weather
We present GPR+ consisting of 7 different types of weather, e.g. sunny, clouds, overcast, foggy, neutral, blizzard and snowlight.
Illumination
We present GPR+ consisting of 7 different types of illumination, e.g. midnight (23:00-4:00), dawn (4:00-6:00), forenoon (6:00-11:00), noon (11:00-13:00), afternoon (13:00-18:00), dusk(18:00-20:00) and night (20:00-23:00).
Background
Various backgrounds of GPR+, including urban areas and wild scenes, such as street, mall, school, park, beach and mountain, etc.
Locations
The position of each location in GTA5 world, in particular, our locations are mainly concentrated in the urban areas and wild scenes, such as street, school, mall, beach, parkding area, etc.
Acknowledgements
We sincerely thank the outstanding annotation team for their excellent work.
This work is partially supported by the National Natural
Science Foundation of China under Project(Grant No.61977045)
and SJTU-SMARCHIT Joint Laboratory of Smart Building.
The template of this webpage is borrowed from Dian Shao.
