About me

Hi, I’m Ganlin Zhang (张甘霖), a PhD student at Technical University of Munich, supervised by Prof. Daniel Cremers. Currently I focus on Visual SLAM, Structure from Motion and 3D reconstruction.

Previously, I received my Master’s degree in Computer Science from ETH Zurich, where I worked on 3D Vision research projects with Prof. Luc Van Gool and Prof. Marc Pollefeys. Before that, I obtained my Bachelor’s degree in Computer Science from ShanghaiTech University, supervised by Prof. Laurent Kneip. During my undergraduate studies, I also spent a year at UC Berkeley, as a visiting student.

News

• 2024.08: I have joined TUM Computer Vision Group as a PhD student, supervised by Prof. Daniel Cremers!
• 2023.03: 🎉🎉 My first first-author paper "Revisiting Rotation Averaging: Uncertainties and Robust Losses" has been accepted by CVPR 2023! Thanks to my advisors and coauthors Dr. Viktor Larsson and Dr. Dániel Béla Baráth for the huge amount of helps. The code has already been available in Github.

Publications

GlORIE-SLAM: Globally Optimized RGB-only Implicit Encoding Point Cloud SLAM Ganlin Zhang*, Erik Sandström*, Youmin Zhang, Manthan Patel, Luc Van Gool, Martin R. Oswald Preprint on ArXiv, 2024 Github Repo | ArXiv | Project website 1. A monocular SLAM pipeline with deformalbe neural pointcloud scene representation. 2. Novel DSPO layer for BA, which can jointly optimize depth map, depth scale and camera pose.

Revisiting Rotation Averaging: Uncertainties and Robust Losses Ganlin Zhang, Viktor Larsson, Daniel Barath Conference on Computer Vision and Pattern Recognition (CVPR), 2023 Github Repo | ArXiv 1. Better model the underlying noise distributions by directly propagating the uncertainty from the point correspondences into the rotation averaging. 2. Integrate a variant of the MAGSAC++ loss into the rotation averaging, instead of using the classical robust losses.

Selected Projects

Holo-Spot: Accessible Robot Control in Mixed Reality Ganlin Zhang*, Deheng Zhang*, Longteng Duan*, Guo Han* Course project of Mixed Realiy 2022 in ETH Zurich Github Repo | Project website In this project, we design, implement and deploy a mixed-reality-based method with HoloLens 2 that enables users to control the Boston Dynamics Spot robot.

NICE-SLAM with Adaptive Feature Grids Ganlin Zhang, Deheng Zhang, Feichi Lu, Anqi Li Course project of 3D Vision 2022 in ETH Zurich Github Repo In this project, we present a sparse version of NICE-SLAM, which is a SLAM system incorporating the idea of Voxel Hashing into NICE-SLAM framework. Instead of initializing feature grids in the whole space, voxel features near the surface are adaptively added and optimized.

Optimization by Particle Swarm Using Surrogates via Bunch-Kaufman Pivoting and Standard Optimization Ganlin Zhang*, Deheng Zhang*, Junpeng Gao*, Yu Hong* Course project of Advanced System Lab 2022 in ETH Zurich Github Repo Focus on speeding up black-box optimization algorithm OPUS from paper Particle Swarm with Radial Basis Function Surrogates for Expensive Black-box Optimization by Rommel G. Regis. Besides, we implement the speed-up C++ version of Bunch-Kaufman Pivoting.

Improved PSMNet for Deep Stereo Disparity Estimation Ganlin Zhang*, Haokai Pang*, Xinyu Shen*, Yunying Zhu* Course project of Deep Learning 2021 in ETH Zurich Github Repo Combining PSM Net, group-wise corr, dilatedResNet, semantic segmentation information to estimate accurate disparity of stereo image pairs efficiently.

Robot Art: Using Robot Arm to Draw Pictures Ganlin Zhang*, Teng Xu*, Weijie Lyu*, Zhenzhong Tang* Ziyuan Hu* Course project of Introduction to Robotics 2019 in UC Berkeley Github Repo | Project website We design a path-finding algorithm that could generate a path to draw a potrait/character in one stroke. Then we use our self-designed control system to draw these path. This project would be useful for any arm-robots with at least 4 joints.