Rafael Padilha

À propos

I got a Ph.D. from University of Campinas (Unicamp, Brazil), working on machine learning and computer vision methods for chronologically analyzing imagery originated from forensic events (e.g., protests, terrorist attacks, fires) and gathered from social media. Among the project’s goals, I aim to sort images of an event in time, date historical pictures, and verify if the timestamp of a photograph has been manipulated. These are challenging tasks but important considering the era we live in, with widespread misinformation and a huge influx of unorganized information that we experience every day. During my Ph.D., I have also done a research internship at the Multimodal Vision Laboratory at the University of Kentucky (KY, USA).

Additionally, I have a M.Sc. in Computer Science from the University of Campinas, in which I proposed lightweight convolutional neural network architectures for facial verification, specifically designed considering the limitations of smartphones and low-powered devices. The project was done in collaboration with Motorola, resulting in a patent and a verification method that has been deployed to their smartphones. During this research, we collected, sanitized, and annotated multiple datasets related to facial biometrics and spoofing detection for smartphones. Besides both projects, I briefly worked on medical imaging analysis. As an undergraduate, I collaborated with Ph.D. students and researchers in devising machine learning approaches based on hand-crafted feature descriptors to diagnose Diabetic Retinopathy on retinal scans.

I am particularly interested in interdisciplinary projects involving Machine Learning and Computer Vision for Social Good, as well as Fairness and Interpretability in AI algorithms. I am also invested in research topics and problems that have the potential to improve society and people’s lives. I firmly believe that AI is a force of change in our world, and AI researchers have both the potential and the responsibility to change it for the better.