A new bilateral project with the Istanbul Technical University (ITU) was approved in June 2020. The PIs on the Slovenian and Turkish sides are Assoc. Prof. Vitomir Štruc and Prof. Hazim Kemal Ekenel.
The computer vision literature typically focuses on either improving the quality of the low-quality probe images or degrading the quality of the high-quality reference images. Within the proposed bilateral project “Face recognition from Low-Quality images (FaceLQ)”, we will explore a different approach and develop face recognition models that facilitate visual recognition by considering both possibilities within a common approach. By matching images not only in the enhanced, but also in the degraded image domain, we expect to achieve superior performance compared to techniques which only pursue a single possibility.
The objectives of the bilateral FaceLQ project are to:
- develop new image enhancement and degradation models for face recognition that can be applied on real-world low-quality image data,
- improve the recognition performance of face recognition models on low-quality imagery,
- investigate new theory and algorithms for the problem of visual recognition from low-quality images,
- make an impact on the relevant research communities with a project focused on a contemporary problem using state-of-the-art machine learning tools (e.g., deep models),
- disseminate the research results through joint publications at top-tier conferences and journals.