Applications focus on autonomous/self-driving cars, marine vehicles and drones
This two-day short course provides an overview and in-depth presentation of the various computer vision and deep learning problems encountered in autonomous systems perception, e.g. in drone imaging or autonomous car vision. It consists of two parts (A, B) and each of them includes up to 8 one-hour lectures.
Part A lectures (6-8 hours) provide an in-depth presentation to autonomous systems imaging and the relevant architectures as well as a solid background on the necessary topics of computer vision (Image acquisition, camera geometry, Stereo and Multiview imaging, Mapping and Localization) and machine learning (Introduction to neural networks, Perceptron, backpropagation, Deep neural networks, Convolutional NNs).
Part B lectures (6-8 hours) provide in-depth views of the various topics encountered in autonomous systems perception, ranging from vehicle localization and mapping, to target detection and tracking, autonomous systems communications and embedded CPU/GPU computing. Part B also contains application-oriented lectures on autonomous drones, cars and marive vessels (e.g. for land/marive surveillance, search&rescue missions, infrastructure/building inspection and modeling, cinematography).
*The course content and exact lecture topics may vary from the above ones depending on recent advances and will be finalized in consultation with the local organizer
The course will take place on 24-25 August 2020.
The course will take place at KEDEA, 3is Septemvriou – Panepistimioupoli, 54636, Thessaloniki, Greece.
You can find additional information about the city of Thessaloniki and details on how to get the city here.
|08:00 – 09:00||Registration||Registration|
|09:00 – 10:00||Introduction to autonomous systems imaging||Localization and mapping|
|10:00 – 11:00||Introduction in computer vision||Deep learning for object/target detection|
|11:00 – 11:30||Coffee break||Coffee break|
|11:30 – 12:30||Image acquisition, camera geometry||Object tracking and 3D locilization|
|12:30 – 13:30||Stereo and Multiview imaging||Parallel GPU and multicore CPU programming. GPU programming|
|13:30 – 14:30||Lunch break||Lunch break|
|14:30 – 15:30||Introduction to neural networks. Perceptron, backpropagation||Fast convolution algorithms|
|15:30 – 16:30||Deep neural networks. Convolutional NNs||Introduction to autonomous drone perception|
|16:30 – 17:30
||Introduction to car vision
|17:30 – 18:30
||Introduction to autonomous marine vehicles|
*Eastern European Summer Time (EEST)
Prof. Ioannis Pitas (IEEE Fellow, IEEE Distinguished Lecturer, EURASIP fellow) received the Diploma and Ph.D. degree in Electrical Engineering, both from the Aristotle University of Thessaloniki, Greece. Since 1994, he has been a Professor at the Department of Informatics of the same University. He served as a Visiting Professor at several Universities. His current interests are in the areas of image/video processing, machine learning, computer vision, intelligent digital media, human-centered interfaces, affective computing, 3D imaging, and biomedical imaging. He is currently leading the big European H2020 R&D project MULTIDRONE. He is also chair of the Autonomous Systems initiative.
Professor Pitas will deliver 12 lectures on deep learning and computer vision.
1. Introduction to autonomous systems imaging
Abstract: This lecture will provide an introduction and the general context for this new and emerging topic, presenting the aims of autonomous systems imaging and the many issues to be tackled, especially from an image/video analysis point of view as well as the limitations imposed by the system’s hardware. Applications on autonomous cars, drones or marine vessels will be overviewed.
2. Introduction in computer vision
Abstract: A detailed introduction in computer vision will be made, mainly focusing on 3D data types as well as color theory. The basics of color theory will be presented, followed by the several color coordinate systems, and finally, image and video content analysis and sampling will be thoroughly described.
3. Image acquisition, camera geometry
Abstract: After a brief introduction to image acquisition and light reflection, the building blocks of modern cameras will be surveyed, along with geometric camera modeling. Several camera models, like pinhole and weak-perspective camera model, will subsequently be presented, with the most commonly used camera calibration techniques closing the lecture.
4. Stereo and Multiview imaging
Abstract: The workings of stereoscopic and multiview imaging will be explored in depth, focusing mainly on stereoscopic vision, geometry and camera technologies. Subsequently, the main methods of 3D scene reconstruction from stereoscopic video will be described, along with the basics of multiview imaging.
5. Introduction to neural networks. Perceptron, backpropagation
Abstract: This lecture will cover the basic concepts of neural networks: biological neural models, perceptron, multilayer perceptron, classification, regression, design of neural networks, training neural networks, deployment of neural networks, activation functions, loss types, error backpropagation, regularization, evaluation, generalization.
6. Deep neural networks. Convolutional NNs
Abstract: From multilayer perceptrons to deep architectures. Fully connected layers. Convolutional layers. Tensors and mathematical formulations. Pooling. Training convolutional NNs. Initialization. Data augmentation. Batch Normalization. Dropout. Deployment on embedded systems. Lightweight deep learning.
1. Localization and mapping
Abstract: The lecture includes the essential knowledge about how we obtain/get 2D and/or 3D maps that robots/drones need, taking measurements that allow them to perceive their environment with appropriate sensors. Semantic mapping includes how to add semantic annotations to the maps such as POIs, roads and landing sites. Section Localization is exploited to find the 3D drone or target location based on sensors using specifically Simultaneous Localization And Mapping (SLAM). Finally, drone localization fusion describes improves accuracy on localization and mapping by exploiting the synergies between different sensors.
2. Deep learning for object/target detection
Abstract: Recently, Convolutional Neural Networks (CNNs) have been used for object/target (e.g., car, pedestrian, road sign) detection with great results. However, using such CNN models on embedded processors for real-time processing is prohibited by HW constraints. In that sense, various architectures and settings will be examined in order to facilitate and accelerate the use of embedded CNN-based object detectors with limited computational capabilities. The following target detection topics will be presented: Object detection as search and classification task. Detection as classification and regression task. Modern architectures for target detection (e.g., RCNN, Faster-RCNN, YOLO, SSD). Lightweight architectures. Data augmentation. Deployment. Evaluation and benchmarking.
3. Object tracking and 3D target localization
Abstract: Target tracking is a crucial component of many vision systems. Many approaches regarding person/object detection and tracking in videos have been proposed. In this lecture, video tracking methods using correlation filters or convolutional neural networks are presented, focusing on video trackers that are capable of achieving real-time performance for long-term tracking on embedded computing platforms.
4. Parallel GPU and multicore CPU programming. GPU programming
Abstract: In this lecture, various GPU and multicore CPU architectures will be reviewed, used notably in GPU cards and in embedded boards, like NVIDIA TX1, TX2, Xavier. The principles of the parallelization of various algorithms on GPU and multicore CPU architectures are reviewed. Sequentially the essentials of GPU programming are presented. Finally, special attention is paid on: a.) fast and parallel linear algebra operations (e.g., using cuBLAS) and, b.) convolution FFS algorithms, as all of them have particular importance in deep machine learning (CNNs) and in real-time computer vision.
5. Fast convolution algorithms
Abstract: Two huge factors related to deep neural network models are the amount of time spend on training such models as well as the response time during DNN inference. Many autonomous systems vision-related applications require very low latency during inference. Both aforementioned are associated with how fast we can compute neural network operations, such as 2D and 3D convolutions. Introducing new and fast ways of conducting these operations can boost the computation speed of deep neural networks.
6.Introduction to autonomous drone perception
7. Introduction to car vision
8. Introduction to autonomous marine vehicles
Any engineer or scientist practicing or student having some knowledge of computer vision and/or machine learning, notable CS, CSE, ECE, EE students, graduates or industry professionals with relevant background.
Registration information will be announced soon.
Remote short course participation is allowed.
Lectures will be in English. PDF slides will be available to course attendees.
A certificate of attendance will be provided.
IF I HAVE A QUESTION?
SAMPLE COURSE MATERIAL & RELATED LITERATURE
1.) C. Regazzoni, I. Pitas, ‘Perspectives in Autonomous Systems research’, Signal Processing Magazine, September 2019
5.) I. Pitas, ‘3D imaging science and technologies’, Amazon CreateSpace preprint, 2019
6.) R. Fan, U. Ozgunalp, B. Hosking, M. Liu, I. Pitas, «Pothole Detection Based on Disparity Transformation and Road Surface Modeling«, IEEE Transactions on Image Processing (accepted for publication 2019).
7.) Rui Fan, Xiao Ai, Naim Dahnoun, «Road Surface 3D Reconstruction Based on Dense Subpixel Disparity Map Estimation«, IEEE Transactions on Image Processing, vol 27, no. 6, June 2018
8.) Umar Ozganalp, Rui Fan, Xiao Ai, Naim Dahnoun, «Multiple Lane Detection Algorithm Based on Novel Dense Vanishing Point Estimation«, IEEE Transactions on Intelligent Transportation Systems, vol. 18, no.3, March 2017
9.) Rui Fan, Jianhao Jiao, Jie Pan, Huaiyang Huang, Shaojie Shen, Ming Liu, «Real-Time Dense Stereo Embedded in A UAV for Road Inspection«, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2019
10.) Semi-Supervised Subclass Support Vector Data Description for image and video classification, V. Mygdalis, A. Iosifidis, A. Tefas, I. Pitas, Neurocomputing, 2017
11.) Neurons With Paraboloid Decision Boundaries for Improved Neural Network Classification Performance, N. Tsapanos, A. Tefas, N. Nikolaidis and I. Pitas IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 14 June 2018, pp 1-11
12.) Convolutional Neural Networks for Visual Information Analysis with Limited Computing Resources, P. Nousi, E. Patsiouras, A. Tefas, I. Pitas, 2018 IEEE International Conference on Image Processing (ICIP), Athens, Greece, October 7-10, 2018
13.) Learning Multi-graph regularization for SVM classification, V. Mygdalis, A. Tefas, I. Pitas, 2018 IEEE International Conference on Image Processing (ICIP), Athens, Greece, October 7-10, 2018
14.) Quality Preserving Face De-Identification Against Deep CNNs, P. Chriskos, R. Zhelev, V. Mygdalis, I. Pitas 2018 IEEE International Workshop on Machine Learning for Signal Processing (MLSP), Aalborg,
Denmark, September 2018
15.) P. Chriskos, O. Zoidi, A. Tefas, and I. Pitas, «De-identifying facial images using singular value decomposition and projections», Multimedia Tools and Applications, 2016
16.) Deep Convolutional Feature Histograms for Visual Object Tracking, P. Nousi, A. Tefas, I. Pitas, IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2019
17.) Exploiting multiplex data relationships in Support Vector Machines, V. Mygdalis, A. Tefas, I. Pitas, Pattern Recognition 85, pp. 70-77, 2019
•Prof. Ioannis Pitas: https://scholar.google.gr/citations?user=lWmGADwAAAAJ&hl=el
•Department of Computer Science, Aristotle University of Thessaloniki (AUTH): https://www.csd.auth.gr/en/
•Laboratory of Artificial Intelligence and Information Analysis: http://www.aiia.csd.auth.gr/