ECCV 2020 Daily - Monday

2 Rudy’s picks DAILY M o n d a y For Today, Monday 24 th Ting-Wu Rudy Chin is a fourth-year Ph.D. student co- advised by Prof. Diana Marculescu and Prof. Gauri Joshi; and a Research Intern at Facebook AI Research (FAIR). My research focuses on efficient deep learning with applications of convolutional neural networks. In short, we are interested in building ConvNets that are more efficient to be deployed or to train. In this ECCV, we have a workshop paper that will be presented at the Embedded Vision Workshop this Friday. Long story very short, we find that if we tune the width of the ConvNets, low-precision networks (weights being quantized to low bit-width) turn out to outperform mixed-precision networks given the same storage cost! Rudy’s picks: • Spotlight: Bounding-box Channels for Visual Relationship Detection ( #4161 ) • Spotlight: Making an Invisibility Cloak: Real World Adversarial Attacks on Object Detectors ( #2294 ) • Poster: Thanks for Nothing: Predicting Zero-Valued Activations with Lightweight Convolutional Neural Networks ( #1027 ) • • Oral: Towards Streaming Perception ( #4349 ) • Oral: End-to-End Object Detection with Transformers ( #4111 ) • Oral: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis ( #4278 ) Find out more on Friday at 12:00pm EST! Read our preview article of this great oral presentation on page 4! ECCV Daily Publisher: RSIP Vision All rights reserved Unauthorized reproduction is strictly forbidden. Designer: Rotem Sahar Our editorial choices are fully independent from ECCV and the conference organizers.