MSR Speaking Qualifier

Title: Model-free Sensorless Manipulation   Abstract: This thesis is a study of 2D manipulation without sensing and planning, by exploring the effects of unplanned randomized action sequences on 2D object pose uncertainty. Our approach uses sensorless reorienting of an object to achieve a determined pose, regardless of the initial pose. Without using sensors and models [...]

Title: OpenPose: Whole-Body Pose Estimation   Abstract: We present the first single-network approach for 2D whole-body (body, face, hand, and foot) pose estimation, capable of detecting an arbitrary number of people from in-the-wild images. Our method maintains constant real-time performance regardless of the number of people in the image. This network is trained in a [...]

Title: Monocular Total Capture: Pose Face, Body, and Hands in the Wild   Abstract: We present the first method to capture the 3D total motion of a target person from a monocular view input. Given an image or a monocular video, our method reconstructs the motion from body, face, and fingers represented by a 3D deformable [...]

Title: 3D Shape Completion and Canonical Pose Estimation with Structured Neural Networks   Abstract: 3D point cloud is an efficient and flexible representation of 3D structures and the raw output of many 3D sensors. Recently, neural networks operating on point clouds have shown superior performance on various 3D understanding tasks, thanks to their power to [...]

Title: Leveraging Simulation for Computer Vision   Abstract: A large amount of labeled data is required to train deep neural networks. The process of data annotation on such a large scale is expensive and time-consuming. A promising alternative in this regard is to use simulation to generate labeled synthetic data. However, a network trained solely [...]

Title: Visual Imitation Learning for Robot Manipulation   Abstract:   Imitation learning has been successfully applied to solve a variety of tasks in complex domains where an explicit reward function is not available. However, most imitation learning methods require access to the robot's actions during demonstration. This stands in a stark contrast to how we [...]

Title: Highly Miniaturized Robots for Inspection of Small Nuclear Piping   Abstract: Bomb making in the 20th century resulted in the creation of massive facilities to produce Uranium. As part of a multi-billion-dollar agenda, the measurement of radioactivity is required for the safe disposal of residual Uranium in piping. Manual techniques have proven too approximate, [...]

Title: Topological Path Planning for Mobile Robot Applications   Abstract: Many path planning problems in mobile robot applications can be solved more efficiently in the topological space. By using the language of topology, the richer spatial information failed to captured by graph/grid-based map representations can be explicitly expressed and exploited. With that, it is possible [...]

Title: Deep Reinforcement Learning with Prior Knowledge   Abstract: Deep reinforcement learning has been applied to many domains from computer games, natural language processing, recommendation systems to robotics.  While model-free reinforcement learning algorithms are promising approaches to learning policies without knowledge of the system dynamics, they usually require much more data. In this thesis, we [...]

Title: 3D Object Detection from CT Scans using a Slice-and-fuse Approach   Abstract: Automatic object detection in 3D X-ray Computed Tomography imagery has recently gained research attention due to its promising applications in aviation baggage screening. The huge dimension of an individual 3D scan, however, poses formidable computational challenges when coupled with deep 3D convolutional [...]