VESSA — Efficient Adaptation of Visual Foundation Models
Self-supervised efficient adaptation of DINO/DINOv2 for specific domains using multi-view video snippets.
Engineer and data scientist focused on machine learning and computer vision. Since 2016, working on R&D with publications and collaborations with Samsung, Petrobras, Google and iFood.
Self-supervised efficient adaptation of DINO/DINOv2 for specific domains using multi-view video snippets.
End-to-end subsurface ML: Seismic segmentation, facies classification, and well-log regression fusing 3D data.
Human Activity Recognition using IMU signals (accelerometer/gyroscope) with on-device optimization.
Inventors: Barreto, J., et al.