Integrated Semantic Visual Perception and Control for Autonomous Systems
Sergiu Nedevschi, TUCN
in the framework of the PN-III-P4-ID-PCCF-2016-0180 SEPCA project
Location: MS-Teams Meeting Link
This workshop presents deep learning based solutions for integrated semantic visual perception and control for Autonomous Systems. The presentations are based on the authors’ experience gain in several research projects carried out at TUCN dealing with perception and control tasks. The approached topics are deep learning based depth perception from single camera, panoptic segmentation, point cloud registration methods and 3D objects detection from point clouds.
1. Depth perception from aerial images using a single camera, Vlad Miclea
2. Panoptic Camera for Environment Perception, Andra Petrovai
3. Comparison of point cloud registration methods, Robert Varga
4. Real-time 3D Object Detection from Point Clouds, Flaviu Vancea
Bosch Students Workshop
Catalin Golban, BOSCH
Location: MS-Teams Meeting Link
By now, it is clear that computer vision based scene understanding will play an extremely important role in the future developments of automated driving technologies. With the current and future technological advancements in computational power, not only in the general-purpose processors and GPUs, but also in the embedded domain, new and more demanding computer vision algorithms are being efficiently deployed to embedded hardware. The biggest challenge now is to be able to accommodate more and more classical and deep learning based solutions on the same embedded system.
The workshop begins with an overview of Bosch Engineering Center Cluj activities in the video-based driver assistance systems and functions and will highlight new algorithmic developments not only in the field of autonomous driving, but also computer vision algorithms for motorcycles, trains and even manufacturing plants.
1. "Introduction and Overview of the Research Topics in Bosch Engineering Center Cluj", Catalin Golban
2. "Stereovision based width clearance", Vlad Stoian
3. "3D segmentation", Andreea Tomoiaga
4. “Direct methods for camera calibration”, Pavel Szabolcs
5. “Measuring scale from surround view cameras”, Andrei Chisca
6. "Computer vision goes to manufacturing (optical inspection for printed labels) ", Andreea Monea
Working Formal Methods Symposium
Faculty of Mathematics and Computer Science of Babes-Bolyai University, the STAR-UBB Institute, the Faculty of Computer Science of the Alexandru Ioan Cuza University of Iasi, ICUB (The Research Institute of the University of Bucharest), and the Faculty of Mathematics and Computer Science of the University of Bucharest.
FROM aims to bring together researchers and practitioners who work on formal methods by contributing new theoretical results, methods, techniques, and frameworks, and/or make the formal methods to work by creating or using software tools that apply theoretical contributions. Formal methods emphasize the use of mathematical techniques and rigour for developing software and hardware. They can be used to specify, verify, and analyse systems at any stage in their life cycle: requirements engineering, modeling, design, architecture, implementation, testing, maintenance and evolution. This assumes on one hand the development of adequate mathematical methods and frameworks and on the other hand the development of tools that help the user effectively apply these methods/frameworks.