Computer Vision |
Publications |
Resources |
CV |
Below is the list of the lectures that are thought at 2nd year MSc:
| Lecture No. |
Topic | What is covered |
| 1 | Introduction to computer vision | history, applications, digital image acquisition, linear algebra for computer vision |
| 2 | Image processing in the spatial domain | kernels, convolution, image derivatives, histograms, histogram equalization |
| 3 | Image processing in the frequency domain | Fourier transform, frequency spectrum, convolution theorem, fast fourier transform, filtering in the frequency domain: low pass, high pass, band pass |
| 4 | Image segmentation | Edges, active contours, Hough transform |
| 5 | Morphological image processing | Otsu's therhold, dilation, erosion, opening, morphological gradient, morphological filters |
| 6 | Colour spaces and image features |
Colour models, converting between colour spaces, colour slicing, colour complements, colour segmentation. Image features: shape, geometrical, statistical, moment-based, geometrical |
| 7 | Image texture classification |
Gray level co-occurence matrix, autocorrelation, texture filters. Introduction to classification algorithms for computer vision: probabilistic modelling: Bayes classifier, K-nearest neighbor, logistic regression |
| 8 | Support Vector Machines and Histogram of Gradients |
Support Vector Machines (SVM) for classification using various kernels Steps of Histograms of Gradients (HoG) feature extraction |
| 9 | Object detection and AdaBoost |
Boosting, AdaBoost algorithm Viola and Jones face detection algorithm |
| 10 | Convolutional Neural Networks |
Computational graph, backpropagation. Deep neural networks: convolutional layers, activation maps, activation functions, subsampling |
| 11 | Transfer Learning and Video Tracking |
Well known CNN topologies Transfer learning Video tracking: mean shift algorithm |
| 12 | Optical Flow | Optical flow equation, frame derivatives, Lukas & Kanade method |
| 13 | Stereo Vision | Disparity, Intrinsic matrix, Reconstruction Error, Rectification |
| 14 | Popular deep learning architectures for Computer Vision | Super-resolution, V-Net |
For any questions, please contact me at petru.radu@e-uvt.ro