High-Resolution Facial Portrait Generation System based on Adversarial CLIPs
Xijun Lu, Chengde Lin · 2024
The portrait of a suspect is a crucial clue in the crime investigation process. Face simulation portrait technology is capable of creating a suspect’s portrait using eyewitnesses’ description of the suspect’s face. Traditional methods suffer from issues of inefficiency, inaccuracy, and lack of realism. To address these issues, we have developed a high-resolution face simulation portrait system utilizing adversarial CLIPs. The system is capable of synthesizing a face that accurately matches the provided description of the suspect’s facial features. Subsequently, these faces are reconstructed at a high-resolution level. The system is deployed on the web, enabling police officers to access high-resolution, authentic face portraits online.