Synthesis of a Cascaded Face Sketch in Various Illuminations PROJECT TITLE : Cascaded Face Sketch Synthesis Under Various Illuminations ABSTRACT: Digital entertainment relies heavily on face sketch synthesis from a photo. A system for intelligent face sketch synthesis must be extremely resilient to changes in lighting. Even in situations when the illumination isn't precisely controlled, such a system is capable of consistently good performance under a wide range of lighting circumstances. Synthesizing a face from a sketch has traditionally been done in a lab with carefully calibrated lighting. Using these procedures can result in disappointing outcomes if the lighting conditions change. A cascaded low-rank representation and a multiple feature generator are used in this paper to propose a new cascaded face sketch synthesiser. Using the multiple feature generator, an artist's drawing style can be replicated and a photo feature can be extracted that is resistant to different lighting conditions. Both capabilities make it possible to select the best possible sketch candidates from the database based on a particular photo patch. In order to reduce the gap between the synthesised face sketch and its artist-drawn counterpart, a low-rank representation of the face is used. Results from the Chinese University of Hong Kong face sketch database show that the proposed cascaded framework produces realistic sketches on par with the current approaches under well-controlled illuminations. Compared to conventional methods, this framework performs significantly better on the extended Chinese University of Hong Kong face sketch library and on Chinese celebrity face photographs from the web under different illuminations. Face sketch synthesis and optical imaging require computer-aided optical systems, and we believe that this framework opens up new possibilities for their application. Did you like this research project? To get this research project Guidelines, Training and Code... Click Here facebook twitter google+ linkedin stumble pinterest The CANet Cross-Disease Attention Network for Diabetic Macular Edema Grading and Joint Diabetic Retinopathy Multi-Instance Learning and the Extreme Value Theorem are used to classify volumetric images.