北京时间7月6日(本周四)9:30-11:30,将门-TechBeat人工智能社区参与的系列直播活动「AI论文背后的故事」第八期——三维视觉: 传统与现代-CVPR Best Student Papers 背后的故事将在视频号准时开播,欢迎预约观看!本活动由香港中文大学(深圳)理工学院、GAP-Lab、未来智联网络研究院组织策划,本期邀请来自Simon Fraser University的Zhiqin Chen,同济大学的陈涵晟和西北工业大学的张曦予一起分享科研背后的故事。AI论文背后的故事主题:三维视觉: 传统与现代-CVPR Best Student Papers 背后的故事时间:2023年7月6日(周四)9:30 - 11:30简介:本期我们有幸邀请了CVPR 2020、2022、2023最佳学生论文的三位一作学生,将围绕传统三维视觉与图形学经典问题和如今基于神经网络的技术如何融合以及如何演进进行探讨。三位嘉宾中有博士即将毕业,有即将开始读博,也有刚开始进行科研的,让我们一起听听他们科研背后的故事。讲者:Zhiqin Chen, Simon Fraser University陈涵晟,同济大学张曦予,西北工业大学主持人:韩晓光教授,香港中文大学(深圳)罗忠金,香港中文大学(深圳)在读博士生组织机构:香港中文大学(深圳)理工学院GAP Lab未来智联网络研究院报告1主题:Revisiting Classic Graphics Methods with Modern Deep Learning Tools讲者:Zhiqin Chen, Simon Fraser University摘要:While classic graphics methods have undergone extensive research and are difficult to be pushed further, they can be very inspirationalin a deep learning context. In this talk, I will briefly introduce how most of my works are derived by combining classic methods, such as implicit surfaces, BSP-trees, image quilting, eigenfaces, marching cubes, dual contouring, andmesh rasterization, with modern deep learning tools.讲者简介Zhiqin is a final-year Ph.D. student at Simon Fraser University, supervised by Prof. Hao (Richard) Zhang. He received his Master's degree from Simon Fraser University and Bachelor's degree from Shanghai Jiao Tong University. He won the best student paper award at CVPR 2020 and best paper award candidate at CVPR 2023. He was an NVIDIA graduate fellowship finalist and received Google PhD Fellowship in 2021. He has also interned at Adobe, NVIDIA, and Google in the past. His research interest is in computer graphics with a specialty in geometric modeling, machine learning, 3D reconstruction, and shape synthesis.报告2主题:End-to-end 3D Vision and Graphics讲者:陈涵晟,同济大学摘要:End-to-end model training has shown immense potential in the field of modern deep learning. This talk will explore the story behind the CVPR 2022 best student paper EPro-PnP and our recent preprint SSDNeRF, both of which originate from a simple idea: end-to-end training of complex 3D vision/graphics models that were formerly managed in two stages. These investigations underscore the superiority of theory-driven loss functions over their empirical counterparts, suggesting the potential of enhancing off-the-shelf models through integrated end-to-end training paradigms.讲者简介陈涵晟,同济大学2023届硕士毕业生(上海市优秀毕业生),斯坦福大学计算机科学博士预录取生。他的研究兴趣为3D计算机视觉和图形学,主要方向包括3D生成与重建、神经渲染、几何视觉等。硕士期间先后在阿里巴巴达摩院,加州大学圣地亚哥分校(远程)实习,在CVPR发表一作论文两篇,获CVPR 2022最佳学生论文奖,同济大学追求卓越学生提名奖,WAIC青年优秀论文提名奖。近期担任CVPR、ICCV、SIGGRAPH、TPAMI、TCSVT等会议和期刊审稿人。报告3主题:3D Registration with Maximal Cliques讲者:张曦予,西北工业大学摘要:三维点云配准是计算机视觉中的一个基本问题,旨在寻找最佳姿态以对齐点云对。本文提出了一种基于极大团(MAC)的三维配准方法,目的是放松先前的最大团约束,并在图中挖掘更多的局部一致性信息以进行准确的姿态假设生成。在多种模态数据集上进行的广泛实验,证明MAC有效地提高了配准精度,优于各种最先进的方法。该方法也能作为一个通用模块插入到深度学习方法中,并能显著提高其性能。我们的工作证明了传统几何方法针对三维点云配准重建仍有巨大的潜力。讲者简介西北工业大学计算机学院在读硕士,主要研究领域为三维点云特征匹配与点云配准,导师为杨佳琪副教授。已在IEEE TPAMI 2023(影响因子24.314,泛AI影响力最高期刊)上发表论文《Mutual Voting for Ranking 3D Correspondences》,在CVPR 2023投稿的论文《3D Registration with Maximal Cliques》被评为最佳学生论文(会议设立该奖的二十三年以来,历史上首次由所有作者均来自国内高校的论文获得)。About GAP LabThe GAP Lab is based on School of Science and Engineering (SSE) and Future Network of Intelligence Institute (FNii) at The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), led by Dr. Xiaoguang Han. GAP means Generation and Analysis of Pixels, Points and Polygons, we are aiming to identify and bridge the research gaps, in the area of generalizing and analyzing images, videos and 3D content (e.g., points and meshes). The research field covers computer vision, computer graphics and machine learning. We are also doing applied research on intelligent medical image analysis and image-based weather forecasting. For more details, please refer to https://gaplab.cuhk.edu.cn/-The End-扫码观看!本周上新!“AI技术流”原创投稿计划TechBeat是由将门创投建立的AI学习社区(www.techbeat.net)。社区上线480+期talk视频,2400+篇技术干货文章,方向覆盖CV/NLP/ML/Robotis等;每月定期举办顶会及其他线上交流活动,不定期举办技术人线下聚会交流活动。我们正在努力成为AI人才喜爱的高质量、知识型交流平台,希望为AI人才打造更专业的服务和体验,加速并陪伴其成长。投稿内容// 最新技术解读/系统性知识分享 //// 前沿资讯解说/心得经历讲述 //投稿须知稿件需要为原创文章,并标明作者信息。我们会选择部分在深度技术解析及科研心得方向,对用户启发更大的文章,做原创性内容奖励投稿方式发送邮件到
[email protected]或添加工作人员微信(chemn493)投稿,沟通投稿详情;还可以关注“将门创投”公众号,后台回复“投稿”二字,获得投稿说明。关于我“门”将门是一家以专注于发掘、加速及投资技术驱动型创业公司的新型创投机构,旗下涵盖将门创新服务、将门技术社群以及将门创投基金。将门成立于2015年底,创始团队由微软创投在中国的创始团队原班人马构建而成,曾为微软优选和深度孵化了126家创新的技术型创业公司。如果您是技术领域的初创企业,不仅想获得投资,还希望获得一系列持续性、有价值的投后服务,欢迎发送或者推荐项目给我“门”:
[email protected]点击右上角,把文章分享到朋友