Computer Vision Methods

Schedule on FEL Students of the course Upload system Discussion forum Labs

Course Description

This course is devoted to computer vision problems: Finding of correspondences between images using image features and their robust invariant descriptors, features matching, picture stitching, object and segment recognition in pictures or video, image retrieval and object tracking in the video sequences.

Pre-requisites

The computer vision methods course expects programming skills in MATLAB computing environment. The programing assignments solving various computer vision methods are a substantial part of the labs. The attendee is expected to know basics of digital image processing as convolution, filtration, intensity transformations, image function interpolations and basic geometric transformations of the image (see the first lab). The attended is also expected to govern basics of linear algebra and probability theory.

Lectures: Monday 11:00-12:30, KN:E-126

Lecturers: JM Jiří Matas, JC Jan Čech, OD Ondřej Drbohlav, MS Milan Šulc

week	Date	Lecturer	Slides	Topic
1	19.2.	JC	Correspondence I-III (s. 1-55)	Interest point and distinguished regions detection: Harris operator (corner detection), Laplace operator and its approximation by difference of Gaussians, affine covariant version, Maximally Stable Extremal Regions (MSER).
2	26.2.	JM	Correspondence I-III (s.56-90)	Descriptors of measurement regions: SIFT (scale invariant feature transform), shape context. Local affine frames for geometric and photometric invariance of description.
3	5.3.	JM	Correspondence I-III (s. 90-124)	SIFT variants, LBP (local binary patterns), Finding correspondences and object recognition using local invariant description.
4	12.3.	JM	RANSAC retrieval, Minhash	RANSAC. Image Retrieval for large image collections: image description, indexing, geometric consistency.
5	19.3.	JC	Deep learning	A shallow introduction into the deep machine learning.
6	26.3.	JC	Deep learning II	Deep learning for object detection. Further insights into the deep nets.
7	2.4.		Easter Monday.
8	9.4.	JM	Tracking I-III mean shift	Tracking I. Introduction. Mean Shift
9	16.4.	JM	Tracking I-III KLT, KCF Tracking	Tracking II. KLT tracker, KCF Kernel Correlation Filter.
10	23.4.	JC	TLD, Tracking_by_Segmentation, Dense_Correspondences	Long-term Tracking TLD: Tracking-Learning-Detection, Tracking by Segmentation, Dense Motion - Optical Flow, Scene Flow
11	30.4.	JM	Viola-Jones face detector, Waldboost	Object detection by sliding window and sequential decision making (Method of Viola and Jones, Waldboost)
12	7.5.	MS		Case study: Plant recognition using Deep Netsb
13	14.5.			Case study: Applications of Visual recognition. Technical specifications and acceptance protocol. Data Collection. Installation. Maintenance
14	21.5.	JM	Hough Transform	Detection of geometric primitives (lines, circles, elipses, etc.). Hough transform and its comparison with RANSAC(Random Sample Consensus).

Evaluation

Work during the semester 50%, written part of the exam 40%, oral part of the exam 10%

Exam

Examples of exam questions. There will be 4-5 similar questions at the written part of the exam. The oral part of the questions takes place after the written part and will focused on discussion of your answers.

Informations

Further informations are available in next sections of this page, we would appreciate your feedback on content and organization on the discussion forum of the course.

Good luck to all participants of the course.

Lecturers

Jiří Matas	Jan Čech	Ondřej Drbohlav

Table of Contents