This page contains the following demonstrations of the Geocomp research group:

Press on the images to view the demos or demo groups for each topic!

Demos for selected publications

Demonstrations connected to selected publications of the GeoComp group. Complete publication list of the group is here.

MVPCC-Net: Multi-View Based Point Cloud Completion Network for MLS Data (2023): This new method comprises a novel multi view-based approach for completing high-resolution 3D point clouds of partial object shapes obtained by mobile laser scanning (MLS) platforms. Code & Benchmark available.

Published in: Image and Vision Computing, 2023 , presented in ICPR 2022.

ST-DepthNet: A spatio-temporal deep network for depth completion using a single non-repetitive circular scanning Lidar (2023): novel a novel depth image completion technique based on sparse consecutive measurements of a
non-repetitive circular scanning (NRCS) Lidar. Code & Benchmark available.

Published in: IEEE Robotics and Automation Letters, 2023

RangeMRF: Point Cloud Based Urban Scene Analysis (2022): This new method comprises 3D point cloud registration and change detection through fusing Lidar point clouds with significantly different density characteristics. Code & Benchmark available.

Published in: Int J Appl Earth Obs Geoinf, 2022 , presented in ICIAR 2020.

ChangeGAN: change detection in coarsely registered point clouds (2021): We introduced a novel change detection approach called ChangeGAN for coarsely registered point clouds in complex street-level urban environment. Benchmark data available.

Published in: IEEE Robotics and Automation Letters, 2021

Deep Learning-based Masonry Wall Image analysis (2020): end-to-end segmentation and inpainting algorithm for masonry images, automatic detection and virtual completion of occluded or damaged regions. Webapp and benchmark available.

Published in: Remote Sensing, 2020 , presented in: ICPR 2021 and ICIAR 2019

HierarchyNet: Hierarchical CNN-Based Urban Building Classification (2020) a new hierarchical deep neural network for the automatic classification of urban buildings from different main and subcategories based on photos of their facades.

Published in: Remote Sensing, 2020

Automatic Camera and Lidar Calibration (2020): a novel automatic, online and target-less camera-Lidar extrinsic calibration approach, for sensor fusion and multi-object tracking from a moving platform.

Published in: Remote Sensing, 2020 , presented: IEEE ICIP 2019, and IEEE MVA 2019

3D CNN Based Semantic Labeling for Mobile Laser Scanning Data (2019): a new method to segment MLS point clouds into nine different semantic classes, which can be used for high definition city map generation. Benchmark available.

Published in: IEEE Sensors Journal, 2019

Lidar-based biometric gait recognition and activity detection (2018) in multi-pedestrian environments (available for Velodyne HDL64 and VLP16 sensors). Benchmark available.

Published in: IEEE Trans. Circcuits and Systems for Video Technology, 2018

Embedded Marked Point Process Framework (2017) a new framework which extends conventional Marked Point Process models enabling multi-level object population analysis. Evaluation is given for three applications. Benchmark is available.

Published in: IEEE Trans. Image Processing, 2017

Instant Lidar object detection (2017): an integrated object detection and classification method using deep learning techniques for extracting various object categories in streamed LIDAR point clouds (cover page article!). Benchmark available.

Published in: IEEE Geoscience and Remote Sensing Letters, 2017

Aerial Lidar-based traffic monitoring (2015): automatic detection of parking and moving vehicles in aerial LIDAR point clouds, and identifying groups of corresponding vehicles, such as cars in a parking lot, or a vehicle queue in front of a traffic light.

Published in IEEE Trans. Geoscicence and Remote Sensing, 2015

Lidar-based long term person tracking with re-identification (2014): a novel real-time surveillance approach working on Velodyne HDL-64 measurement sequences..

Published in Pattern Recognition Letters, 2014

ISAR Object Detection (2014): Target Detection, Recognition and Tracking in Inverse Synthetic Aperture Radar (ISAR) Images

Published in IEEE Trans. Geoscicence and Remote Sensing, 2014

Automated Optical Quality inspections of Printed Circuit Boards (2013): research work conducted in cooperation with the Dept. of Electronics Technology, Budapest University of Tech. and Economics.

Published in IEEE Trans. on Industrial Electronics, 2013 and in PRL, 2011

Multi Camera People Localization and Height Estimation (2013).

Published in IEEE Trans. on Circuits and Systems for Video Technology, 2012 presented in CVPR 2011 .

Building Extraction and Change Detection (2012) in Multitemporal Aerial or Satellite Images based on Marked Point Processes. Benchmark data is available.

Published in IEEE Trans. Pattern Analysis and Machine Inttelligence, 2012

Change detection in optical aerial images (2009) with large time differences. Benchmark dataset is available

Published in IEEE Trans. Geoscicence and Remote Sensing (2009)

Object motion detection from a moving aerial platform (2009) by a a three-layer MRF model (2006-2009). Benchmark dataset is available

Published in IEEE Trans. Image Processing (2009)

Foreground and Shadow Detection in Surveillance Videos (2008). Visit the bechmark database set web page.

Published in IEEE Trans. Image Processing (2008) and ACCV 2006.

 

Demos for featured projects

Demonstrations connected to featured projects of the Geocomp group. Complete list of present and past projects of the group is here.

Ready2BIM Demonstrator (2022): browser independent application to measure and analyze facade elements of buildings. In the application point clouds of the facades can be uploaded, and various measurements can be applied that can help place insulation elements on them.

i4D film preproduction system. Aim: supporting a complete storyboard designing workflow by enabling one to synthesize and visualize virtual 3D scenes from real and fictional elements.

Further demos on Lidar based road scene understanding (2012-2016): we developed automatically interpret the LIDAR point cloud stream obtained from a moving platform, segment different point cloud classes, detect and recognize various field objects

Processing measurements of mobile and terrestrial mapping systems (MLS/TLS): we develop automated algorithms for point cloud cleaning, monitoring various public premises, including road quality assessment, surveys of road marks and traffic signs, urban green area estimation, traffic analysis

i4D Project: Virtual city modeling and moving person/crowd visualization (2012-2014): building facade and roof model reconstruction, texture polygon or colored point cloud visualization.

 

i4D Project: 4D video surveillance (2012-2014) we jointly exploit various depth sensors and optical cameras, to achieve high level object detection and classification, multiple people localization and tracking, event recognition a biometric identification tasks.

4D people model generation (2010-2013): the dynamic models of walking pedastrians are prepared in the 4D Reconstruction Studio of SZTAKI


    


Links

Geo-Information Computing @ Machine Perception Lab.

GeoComp Research Group

Machine Perception Laboratory

SZTAKI main page

GeoComp Demos:

Demo page

Publications

Contact

GeoComp Group leader: Prof. Csaba Benedek benedek.csaba@sztaki.hu

i4D project manager: Dr. Zsolt Jankó janko.zsolt@sztaki.hu

Head of MPLab: Prof. Tamás Szirányi

MPLab administration: Anikó Vágvölgyi

Address:

SZTAKI
Kende utca 13-17
H-1111 Budapest, Hungary
Tel: +36 1 279 6097

Flag Counter