Szeliski, "Computer Vision: Algorithms and Applications", 2010.
Koenderink, "Solid Shape", MIT Press, 1990 Zisserman, "Multiple View Geometry in Computer Vision", Cambridge University Press, 2004. Papadopoulo, "Geometry of Multiple Images", MIT Press, 2001. Zisserman "Toward Category-Level Object Recognition", Lecture Notes in Computer Science 4170, Springer-Verlag, 2007 Ponce, "Computer Vision: A Modern Approach", Prentice-Hall, 2nd edition, 2011 High level video analysis, action recognitionĭ.A. Low-level video analysis: tracking, human segmentation and poseĪssignment 4 Due (MNIST recognition with NN) Using synthetic data/3D shape analysis with CNNs Optical Flow: optical flow equation, Lukas-Kanade, Horn and Schunk, SIFT-flow, large displacement and deep optical flow. Introduction to category-level recognition / Introduction to CNNsĭocument on Stochastic Gradient Descent (Guillaume Obozinski)Īnalyzing CNNs, CNNs for object detection and semantic segmentation
Ref: Linear algebra for Vision (from Fei Fei Li): slides, pdf TP2 on mean-shift segmentation ( html, ipynb ) Szeliski chapter 2 "Image formation" + printed introduction Refs : Forsyth and Ponce "Geometric camera model" chapter Human color perception, color in computer vision Markov Random Fields: optimization methods (graph-cuts, belief propagation, TRW-S), applications to stereo and segmentation Instance recognition, Feature detectors and descriptors, SIFT, Visual search
TP on Canny Edges ( html, ipynb, lena.jpg, tools.jpg ) 2 and 3) Detailed presentation of Bilateral Filter ,Įdges (Canny), Segmentation (K-means, GMM, Mean shift), Points (Harris Corners, blob detection) Ressources: relevant Book chapters on Fourier and linear image filtering (chapt. Linear and non-Linear Image filtering: Fourier and convolution, Bilateral Filter, Non-Local-Mean Low level Computer Vision, image correspondences and grouping Introduction, overview, image formation, digital photography If a plagiarism is detected, the student will be reported to ENS. Any uncredited reuse of material (text, code, results) will be considered as plagiarism and will result in zero points for the assignment / final project. The assignments and final projects will be checked to contain original material. However, each student has to work out their assignment alone (including any coding, experiments or derivations) and submit their own report. Discussions are encouraged and are an essential component of the academic environment. You can discuss the assignments and final projects with other students in the class. You will have to present your project (10 minutes + questions) and return a summary (2 pages max) of the essential points that should be readable (and useful) for the other students in the class. This will have to be adapted depending on the paper. You are expected to understand and present the paper, but also to offer some added value, such as experiments of your own, new interesting tests with available code, or comparison with other relevant works.
Feel free to ask for papers on a topic that you are interested in or propose a paper (in this case, it has to be validated before the November 8th)
Each project is based on a paper and a list of suggested papers is available here. The final project will represent 40% of the grade. You have to send your assignments to Robin by the deadline. The supporting materials for the programming assignments projects will be in Python. There will be four/five programming assignments representing 60% of the grade. Robin is your main contact for anything related to the programming assignments and final projects. Teaching Assistant: Robin Champenois, bonjour robin-champenois fr.
TP1 on Canny Edges ( html, ipynb, lena.jpg, tools.jpg ) available, due for October 11th (return by e-mail to Robin).TP2 on mean-shift segmentation ( html, ipynb ), due November 1st.Project choice by November 8th (by e-mail to Robin).TP3 on camera calibration ( html, ipynb ), due November 15th.TP4 on Neural Networks ( html, ipynb ), due December 13th.A cleaner version will be available in January.
Optional TP on optical flow ( html, ipynb ): there might be issues using the code with python3.Send your presentation to the first student of the session (see lecture 15 slide 3). Mathieu Aubry, Karteek Alahari, Ivan Laptev, and Josef Sivic Introduction to Computer Vision 2018/2019