3D Computer Vision
Ongoing
Grant Funded ($1.2M)
OmniVision-3D: Self-Supervised Vision Transformers for Volumetric Scans
Lead Investigator: Dr. Sarah Chen
Executive Summary
Pre-training 3D transformers on 150,000 unannotated volumetric scans for zero-shot tumor detection and organ segmentation.
Technology Stack & Domain Frameworks
Self-Supervised
3D CT/MRI
Transformers
Methodology & Clinical Validation
Project Overview:
OmniVision-3D introduces masked autoencoder pre-training natively designed for 3D spatial grids, dramatically reducing the need for dense voxel manual annotations.