Insights
Detection and recognition are recurring themes in MICC’s history of research and development. In the FREEWAY (Face Recognition and Body Re-identification in the Wild) project we are consolidating our technology and know-how to the Italian multinational aerospace, defense and security company Finmeccanica. The objective of FREEWAY is to develop advanced prototypes of video surveillance systems for deployment In the Wild. Finmeccanica provides the infrastructure and on-site access to large scale, real-world video surveillance installations. MICC provides years of scientific and development expertise in face and person detection, face recognition, and person re-identification.
The FREEWAY project is organized around the themes of:
- Open- and closed-world face recognition In the Wild
MICC is evaluating its previously built laboratory prototypes of systems for face recognition for use in the FREEWAY project. A requirement of the FREEWAY system prototype is the support of both open- and closed-world recognition. For closed-world face recognition, subjects are formally enrolled in the system using example face images. In open-world recognition, unknown subjects are auto-enrolled in the system upon first view. At MICC we are evaluating our face recognition technology for application closed-world recognition in the wild, and our technology for feature hashing and indexing for open-world recognition in the FREEWAY prototype. - Full-body person re-identification In the Wild
In large scale video surveillance scenarios, high resolution images of faces are not always available due to the large area of surveillance covered. For determining short- to medium-term correspondences between persons in an area of surveillance, person re-identification is a technology that can be used to identify previously viewed persons across multiple cameras. Person re-identification is typically performed on full body images, and in the FREEWAY project MICC is investigating the possibility of leveraging our expertise in person re-identification for use in the FREEWAY prototype. We believe that person re-identification can augment the main task of face recognition by instances of the same person across multiple views. - Integration and consolidation
While testing techniques in a laboratory is a mandatory step when developing new effective and efficient systems, on field testing opens new difficulties and scenarios that are hardly reproducible in a controlled environment. Considering this, in the FREEWAY project, MICC and Finmeccanica are cooperating to bring developed prototypes into real world systems. The on field testing consolidates our developing retrieval and annotation technology and permits to obtain a more efficient system, with a lower computational demand.
