DeepFakes detection project receives access to FCT supercomputer

VIS Team has received approval to use the Deucalion platform, part of the National Advanced Computing Network (RNCA) managed by the Foundation for Science and Technology (FCT) and the National Scientific Computing Foundation (FCCN), for an innovative deepfake detection project. The project, led by PhD student Miguel Leão Correia and supervised by Prof. Nuno Gonçalves, aims to create a large dataset using implicit neural representations to distinguish authentic videos from digital manipulations. With the need to process over 130,000 videos from the Deepfake Detection Challenge Dataset, high-performance computing resources are critical: without them, the task would require years of computation, rendering scientific deadlines unfeasible.

The RNCA infrastructure, coordinated by the FCT and operated by the FCCN, provides access to advanced processing and storage resources, essential for projects requiring massive computational capacity. The approval will allow the use of 10 GPUs on the Deucalion platform, reducing the estimated processing time from 2,450 days to approximately 158 days. This support not only accelerates the validation of the scientific hypothesis but also positions the study at the forefront of combating deepfakes, a growing threat on social media. The initiative reinforces the RNCA’s strategic role in driving research with national and global impact, aligned with the mission of Portugal INCoDe.2030 to strengthen digital competencies and technological innovation.