Update, 14 February 2018: new paper, extended database, evaluation protocol, demo

• Paper published in the IEEE Transactions on Information Forensics and Security, entitled ‘Matching Software-Generated Sketches to Face Photos with a Very Deep CNN, Morphed Faces, and Transfer Learning‘, describing a method for face photo-sketch recognition using deep learning as also detailed here. The method is evaluated primarily on the software-generated sketches in the UoM-SGFS database, which has been doubled in size.
• A simple demonstration of the method above is also available to download, where the identity of a subject in a sketch is found by comparing it to 3 photo images. More details here.
• As mentioned above, the UoM-SGFS database has been doubled in size to now contain 1200 images of 600 subjects. To the best of our knowledge, this currently makes the database (i) one of the largest face sketch databases, (ii) the largest face sketch database containing software-generated sketches, (iii) the only database containing all sketches represented in full colour, (iv) the only database containing sketches created with EFIT-V, and (v) one of the few databases containing more than one sketch image per subject. More details here.
• The protocols used to split the UoM-SGFS database into training and testing sets in the paper ‘Matching Software-Generated Sketches to Face Photos with a Very Deep CNN, Morphed Faces, and Transfer Learning‘ and in ‘Forensic Face Photo-Sketch Recognition Using a Deep Learning-Based Architecture‘ have also been made available here. This allows algorithms to be directly compared with the methods evaluated in these two papers, serving as a benchmark for future algorithms.