When face recognition fails, person re-identification is needed.
With person re-identification, we aim to determine whether a given person has already appeared over a network of cameras. A robust modelling of the appearance of the person is necessary for succesful re-identification. This is a challenging task as significant appearance changes may be caused by variations in lighting conditions, camera view angles and different poses of the person.
Furthermore, attribute recognition helps to identify the characteristics of the target person. The attribute could be gender of the person, clothes worn by the person, accessories carried by the person, etc.
The person re-identification and human attribute recognition features have been developed using state-of-the-art Deep Learning technology.
We achieved top international performance in these technologies based on publicly available datasets.
These technologies equip security personnels with better and faster capability to automatically locate, search and identify a target person via a network of surveillance cameras, which otherwise would require manual inspection.
With higher accuracy in person re-identification and human attribute recognition, we can realize a better surveillance system.
Person re-identification and human attribute recognition are important features for next-generation surveillance systems, to meet the needs of an increasingly complex security landscape.
Beyond surveillance, the technologies may be utilized in commercial, marketing or various other areas.
Higher accuracy of Person Re-Identification and Human Attribute Recognition system leads to a better surveillance system.
Customer can enjoy more comprehensive surveillance solution.