MICrONS (Machine Learning from Cortical Networks) is scientific research program funded by IARPA with the mission to revolutionize machine learning by understanding the detailed structures and functions of mesoscale neural circuits, and then exploiting the mathematical functions underlying the algorithms of the brain. It is an Apollo project of the brain.
Develop and apply deep learning algorithms to screen for pancreatic neoplasms using CT and MR imaging. Funded by the Lustgarten Foundation.
Goals/Objectives: (1) Design and learn appearance models that stitch together to create a generative pixel-level image model. (2) learn compositional models of objects. Current state-of-the art computer vision models are bottom-up, discriminative, and non-generative. We will develop a new class of generative three-dimensional models of object which directly visual appearance at the image pixel level. This will lead to interpretable models that correctly weight evidence during inference. We have shown that sufﬁcient statistics can be learned and used in a conditional modeling framework to build complex, realistic, appearance models from low-dimensional distributions. This will be combined within a hierarchical compositional framework. This approach to hierarchical generative models will enable us to effectively parse three-dimensional objects and parts. This will result in practical computer vision systems which can parse their environment and perform a large set of visual tasks. N00014-15-1-2356.
The human vision system understands and interprets complex scenes for a wide range of visual tasks in real-time while consuming less than 20 Watts of power. This Expedition in Computing project explores holistic design of machine vision systems that have the potential to approach and eventually exceed the capabilities of human vision systems. This will enable the next generation of machine vision systems to not only record images but also understand visual content. Such smart machine vision systems will have a multi-faceted impact on society, including visual aids for visually impaired persons, driver assistance for reducing automotive accidents, and augmented reality for enhanced shopping, travel, and safety. The transformative nature of the research will inspire and train a new generation of students in inter-disciplinary work that spans neuroscience, computing and engineering disciplines. NSF award CCF-1317376.
Our scientiﬁc goal is to discover how intelligence is grounded in computation, how these computations are implemented in neural systems, how they develop during childhood, and how social interaction ampliﬁes the power of these computations. As we progress, we will aggressively pursue opportunities to discover and develop unifying mathematical theories. To foster collaboration across disciplines, we will jointly develop top-to-bottom computational models powerful enough to explain visually perceived situations the way humans do. The models will emerge from fundamental questions about visually perceived situations: who, what, why, where, how, with what motives, with what purpose, and with what expectations. Models of visual understanding will be further advanced by developing computational models of what children know and learn about physical objects and intentional agents,and how they learn so much so rapidly. We will develop computational models of learning, memory, reasoning, and concept formation that are consistent with behavior, neural systems, and neural circuits. We will also develop computational models that enable computers to think new thoughts, imagine new scenes, form hypotheses, propose interventions, and compose narratives. Through these collaborative efforts, we will develop new methodologies and new technologies that will help to reach our goals. NSF STC award CCF-1231216.
The objective of this research is to develop computer vision algorithms for jointly segmenting images and detecting the objects within them. The output is a labeling of all the pixels in the image. The aim of this proposal is to develop algorithms which perform much better than the current state of the art, while also developing a framework that can be extended to eventually develop algorithms that perform as well as, or better than, human observers. ARO 62250-CS.