PhD Dissertation Defense Raniero Lara Garduno March 12, 2021

PhD Dissertation Defense

Friday, March 12, 2021                                                         

Virtual Defense via Zoom

                                                                                          

Title: Machine Learning and Digital Sketch Recognition Methods to Support Neuropsychological Diagnosis and Identification of Cognitive Decline

Abstract: With approximately 15 to 20 percent of adults aged 65 and older living with Mild Cognitive Impairment (MCI), researchers in neuropsychology have placed increasing emphasis in early detection to best preserve quality of life. This dissertation presents digital diagnosis tools by adapting existing neuropsychological tests and fully automating what is otherwise a subjective process requiring domain expertise. We present the first fully-automated Rey-Osterrieth Complex Figure grader that can recognize all 18 grading details using a series of agent-based graph traversal algorithms combined with a modified template- matching gesture recognition model. We also present among the first systems to recognize MCI on digitized Trail-Making tests combining machine learning methods with digital sketch recognition.

Biography:  Raniero specializes in the intersection between neuropsychology and digital sketching, more broadly in how subjects' behavior when interacting with digitized examinations. He has studied the effects of cognitive decline on touch tablets, stylus input, digitized paper-and-pencil sketching, and tests integrating augmented reality. He hopes to see technology improve such that Mild Cognitive Impairment can be detected early through the analysis on how people interact with various digital input modalities in test and everyday life.

Advisor: Dr. Tracy Hammond

SRL MS Thesis Defense of Duc Hoang March 2, 2021

MS Thesis Defense

Tuesday, March 2, 2021                                                         

Virtual Defense via Zoom

                                                                                          

                                                             

Title: 3M-Pose: Multi-resolution, Multi-path and Multi-output neural architecture search for bottom-up pose prediction

Abstract: Human pose estimation is a challenging computer vision task and often hinges on carefully handcrafted architectures. This paper aims to be the first to apply Neural Architectural Search (NAS) to automatically design a bottom-up, one-stage human pose estimation model with significantly lower computational costs and smaller model size than existing bottom-up approaches. Our framework dubbed 3M-Pose co-searches and co-trains with the novel building block of Early Escape Layers (EELs), producing native modular architectures that are optimized to support dynamic inference for even lower average computational cost. To flexibly explore the fine-grained spectrum between the performance and computational budget, we propose Dynamic Ensemble Gumbel Softmax (Dyn-EGS), a novel approach to sample micro and macro search spaces by allowing varying numbers of operators and inputs to be individually selected for each cell. We additionally enforce a computational constraint with a student-teacher guidance to avoid the trivial search collapse caused by the pursuit of lightweight models. Experiments demonstrate 3M-Pose to find models of drastically superior speed and efficiency compared to existing works, reducing computational costs by up to 93% and parameter size by up to 75% at the cost of minor loss in performance

Biography:  Duc is an aspiring graduate researcher hailing from Vietnam. He has specialized in the application of Machine Learning in Computer Vision tasks. He is currently pursuing a Master's degree in Computer Science at Texas A&M and later a Ph.D. degree in Electrical andoang Computer Engineering at the University Texas of Austin.

Advisor: Dr. Tracy Hammond