Projects
Applied AI, computer vision, and machine learning engineering projects by Rafael Testa.
My projects span generative video, geometry-aware image synthesis, deepfake detection, and LLM-based analytics. For each project, I describe the problem, the implementation, and the evidence used to evaluate the relevant behavior.
Generative vision · Research
Temporal Coherence for Facial Video Synthesis
Treating temporal coherence as a first-class property in image-to-video facial-expression synthesis.
Approach: Previous-frame conditioning with pixel-transition and facial-landmark consistency objectives.
View GitHub repository ↗
Geometry-aware vision · Research
Facial Expression Synthesis Based on Similar Faces
A geometry-aware computer-vision pipeline for retrieving a similar face and transferring an expression.
Approach: Similar-face retrieval followed by landmark-guided warping and texture transfer.
View GitHub repository ↗
Synthetic media · Reproducible ML
Deepfake Detection Based on Ratio Images
Detecting manipulated videos through interpretable temporal differences between adjacent frames.
Approach: Adjacent-frame ratio images, engineered face and background features, and video-level Random Forest classification.
View GitHub repository ↗
Applied AI · Research engineering
Reliable Natural-Language Analytics
A validation-first architecture for asking questions about public-sector analytical data in natural language.
Approach: Ground, generate, validate, execute with restricted access, then choose a verified presentation.
View project details →