Bio: Dr Chua is the KITHCT Chair Professor at the School of Computing, National University of Singapore. He was the Acting and Founding Dean of the School from 1998-2000. His main research interests fall under the topics of multimedia information retrieval, unstructured multimodal analytics, and the emerging applications in Chabot, wellness and Fintech. He is the Co-Director of NExT, a joint Center between NUS and Tsinghua University on Extreme Search. 

Dr Chua is the recipient of the 2015 ACM SIGMM Achievements Award. He is the Chair of steering committee of ACM international Conference on Mutilmedia Retrieval (ICMR) and Multimedia Modeling (MMM) Conference series. Dr Chua is also the General Co-Chair of ACM Multimedia 2005, ACM CIVR (now ACM ICMR) 2005, ACM SIGIR 2008, and ACM Web Science 2015. He serves in the editorial boards of several international journals. Dr Chua is the Co-Founder of two technology startip companies in Singapore. He holds a PhD from the University of Leeds. More about Dr Chua can be found here.

Title: Revisiting Visual Food Recognition and Beyond

Abstract:

Bio: Daniel is a Senior Research Engineer in Deep Learning at Magic Leap, currently focused on pioneering new learning-based methods for Visual SLAM, co-advised by Tomasz Malisiewicz and Andrew Rabinovich. He received his Master's and Bachelor's degrees at the University of Michigan, where he focused on Machine Learning, Computer Vision and Robotics. During this time he worked on various small projects on topics including person tracking, outdoor SLAM, scene text detection, 3D voxel convnets, robotic path planning, and text summarization, advised by great mentors like Matthew Johnson-Roberson, Edwin Olson, Silvio Savarese and Homer Neal. During his Master's studies he did an internship with Occipital where he helped release the Structure SDK which runs RGB-D SLAM on mobile devices. More about Daniel can be found here.

Title: Learning Deep Convolutional Frontends for Visual SLAM

Abstract:

The self-supervised framework for training interest point detectors and descriptors suitable for a large number of multiple-view geometry problems in computer vision. As opposed to patch-based neural networks, our fully-convolutional model operates on full-sized images and jointly computes pixel-level interest point locations and associated descriptors in one forward pass. We introduce Homographic Adaptation, a multi-scale, multihomography approach for boosting interest point detection repeatability and performing cross-domain adaptation (e.g., synthetic-to-real). Our model, when trained on the MS-COCO generic image dataset using Homographic Adaptation, is able to repeatedly detect a much richer set of interest points than the initial pre-adapted deep model and any other traditional corner detector. The final system gives rise to state-of-the-art homography estimation results on HPatches when compared to LIFT, SIFT and ORB.