The project caught the attention of the evaluation committee primarily due to its potential for immediate practical application. It is a set of forensic analytical tools for image and video processing, designed for use by the criminal police and investigators. Among other things, the project, which utilizes artificial intelligence, has resulted in software that significantly speeds up the investigation of serious sexual crimes.
Automation of terabytes of data checks during investigations
A key part of the solution is the FACIS platform, which automates all steps leading to the discovery of defective content in image data stored on terabyte-sized storage media. According to police experts, this is a task that has not yet been feasible in this form.
Investigating this type of crime is extremely mentally and time-consuming—it requires manual visual inspection of hours of image and video material in a single case. Thanks to the results of the FACIS project, this activity can now be fully automated, and investigators can focus primarily on verifying the relevance of automatically generated outputs.
The FACIS platform was piloted at the Regional Directorate of the Czech Police in Prague, where it received a very positive response and is already being successfully used to solve current cases in Prague and the surrounding area. The application of the project results is being further developed – the Police Presidium of the Czech Republic, the Regional Directorate of the Moravian-Silesian Region, and the Military Police have expressed interest in deploying the platform. Members of the research team are also in contact with interested parties abroad.
Cutting-edge computer vision algorithms from FEE CTU
Experts from the Department of Cybernetics at FEE CTU – Prof. Jiří Matas, Ing. Miroslav Purkrábek, and Ing. Matěj Suchánek from the visual recognition group, and Dr. Vojtěch Franc and Ing. Jakub Paplhám from the machine learning group – participated in the project primarily by developing computer vision algorithms for detecting figures, estimating their age, posture, and other attributes necessary for criminal investigation.
The results of the project were presented at the world's most prestigious conferences in the field, such as Computer Vision and Pattern Recognition (CVPR) and the International Conference on Computer Vision (ICCV). The model for estimating age from a photograph of a face, developed by Vojtěch Franc and his doctoral student Jakub Paplhám, is ranked among the world's best according to the US National Institute of Standards and Technology (NIST). Its error rate differs from the best commercial method by only a few months, and in age categories up to 20 years, it is significantly more accurate than all previously published or open-source solutions.
Winning the Minister of the Interior Award confirms the international level of research at FEE CTU and its direct and measurable contribution to security practices in the Czech Republic.
