NOTE: this course is discontinued as of 2016 and is replaced by 5SH0!
As an introduction, learn theory and practice of the newest advanced video compression standards, like wavelets in JPEG2000, and MPEG-4/H.264. The main part of the course is about video content analysis.
Goal:
Learn the power of signal transformations
- For video compression, like as in transform coding
- For video analysis, like as in the Gabor wavelet transform
- For combining them with Deep Learning
Exploit the concept of “condensing information”
- This is useful for coding and compression
- But also applicable to video analysis: what is the key informative feature
- Use specific transformation for that, like the PCA
Learn the theory and algorithms of video signal analysis, and finding features and objects in image/video. Learn the basics of 3D image processing, sensing in 3D and 3D model reconstruction. Learn the practical basics of MATLAB programming and analysis and/or 3D algorithms for practical applications in surveillance and medical imaging.
Learn the theory and algorithms of video signal analysis, and finding features and objects in image/video. Learn the basics of 3D image processing, sensing in 3D and 3D model reconstruction. Learn the practical basics of MATLAB programming and analysis and/or 3D algorithms for practical applications in surveillance and medical imaging and learn the basics of deep learning.
This course on advanced image/video/3D processing brings you to the level of interesting internships and projects in computer vision at MSc level. The course discusses state-of-the-art techniques in image/video feature analysis, like texture (edge) analysis, motion analysis, 3D reconstruction and object detection/classification. Also, the course introduces you the most efficient deep learning concepts. You will gain a hands-on experience with the computer vision techniques by capturing and processing the 2D/3D imaging data from a UAV drone.
The course starts with the main concept of camera projection matrix and explains the working principles of different imaging sensor types (mono, stereo, ToF, LiDAR, etc) and image data types (RGB, disparity, depth, point cloud, mesh, etc). The video analysis part of the course discusses segmentation techniques and the commonly used feature descriptors: SIFT, HOG, all with the associated filtering. Using these techniques as a basis, we proceed to motion analysis, as well as object detection and tracking techniques. The course also introduces object classification techniques like Gaussian Mixture Modeling, K-means clustering, SVM and basics of Deep Learning. In the 3D imaging part of the course, the 3D multi-view geometry, 3D reconstruction and visual SLAM techniques are presented.
Goal
Become acquainted with the implementation of image filtering in the spatial and frequency domain, the application of image restoration and color image processing algorithms, finding image attributes (points, edges, etc.) and features, controlling segmentation parameters for obtaining optimal results and gain a basic understanding of machine learning algorithms and their validation.
Goal
Become acquainted with the theory and main concepts of digital signal and analysis; make a choice of an approach for a given basic signal/image processing problem; implement the chosen approach in MATLAB and critically evaluate the result; be able to operationalize theoretical concepts from signal/image analysis in a hands-on case study.
WIC Midwintermeeting on Deep Learning TU Eindhoven, Zwarte Doos (Filmzaal), January 24, 2018.
The werkgemeenschap Information and Communication theory organizes annually a one-day meeting with tutorial lectures and new work on an actual technical theme. This year the theme is Deep Learning, which is revolutionizing areas in computer vision, signal processing and machine learning.
