Computer Vision and Image Synthesis
Human vision lets us recognize and interpret vast amounts of information with just a glance. It’s an act so integrated into our brain that we rarely pause to notice how powerful it is. Computer vision is the analogous science by which machines learn to understand and reason with the world from images and video. Not surprisingly, today’s most successful computer vision approaches are biologically inspired and employ deep neural networks – modeled after the human brain – to effectively accomplish a breadth of complex visual tasks, ranging from recognizing rare bird species to detecting early malignancies in bio-tissue.
Computer vision researchers at PARC pursue four complementary avenues of investigation
1. Designing and training deep visual neural networks to work effectively in severely sparse data scenarios by incorporating domain constraints and physical first principles
2. Applying deep neural networks in resource-constrained environments (e.g., mobile or edge platforms) by developing novel abstractions and approximations
3. Enhancing what the machine can ‘see’ via augmented modalities (such as hyperspectral imaging and depth sensing), and incorporating these into a deep inferencing pipeline
4. Developing computational frameworks for machines and humans to collaborate to accomplish complex visual tasks that are too difficult for either entity alone
Our work is centered around a series of Focus Areas that we believe are the future of science and technology.
We’re continually developing new technologies, many of which are available for Commercialization.