Computer'Vision Course'Introduction Prof.&Flávio&Cardeal&– DECOM&/&CEFET7MG cardeal@decom.cefetmg.br
Who$am$I? • Academic)Background o D.)Sc.)4 Computer)Science,)UFMG. o M.)Sc.)4 Computer)Science,)UFMG. o BEng.)4 Electrical)Engineering,)TU4Berlin/UFMG. • Research)Interests o Computer)Vision. o Image/Video)Processing. • Associate)Professor) o Department)of)Computing)4 CEFET4MG. o Homepage:)http://cardeal.piim4lab.cefetmg.br. 2
Why$EMI$in$this$course? • English Medium Instruction (EMI). • According to UNESCO, 4.5 million students study outside their country of citizenship. • It is estimated that the number of internationally mobile students could reach 7 million by 2020. • In this context, university managers believe that their institutions have to internationalize in order to progress in the world rankings . 3
Why$EMI$in$this$course? • In order to become an international university the institution has to attract students from other countries. • Those students usually do not speak the language of the host country. • Therefore the language of instruction has to be one that all students will understand and English has demonstrated to be a good choice. 4
Why$EMI$in$this$course? • Moreover, in order to internationalize, the institution has to demonstrate that its teaching and research force, its faculty , is multinational . • People believe that multinational is synonymous with better quality than mono=national. • Finally, EMI is considered an authentic way to learn a language, by addressing situations which are encountered in reality. 5
Why$EMI$in$this$course? • So, the world is experiencing a rapid increase in the teaching of subjects through EMI in countries where the first language is not English. • In this scenario, by using EMI in our course, I expect to contribute to include our institution in this growing global phenomenon. 6
Course'Description • Introduction to basic concepts in Computer Vision , a research field that develops methods for machines to understand images/videos . • We will explore topics that contribute to deal with problems such as the following ones. 7
Course'Description • Can we interact with machines in richer ways, perhaps with gestures or facial expressions? Source:)Leap)Motion)Inc. 8
Course'Description • How can a self,driving vehicle identify objects in complex environments? Source:)Google)Inc. 9
Course'Description • How can a camera in the operating room help a surgeon plan a procedure more safely? 10 Source:*Gemelli*Hospital*
Course'Description • Given billions of images, how can you find one that “looks like” some image of interest? Source:*Mad*Magazine 12
Prerequisites • This course is intended for graduate and upper)level undergraduate students. • Linear algebra, elementary statistics and a programming language (e.g., Python or C). 13
Objectives • To understand the concepts, problems, and solution techniques in computer vision. • To apply computer vision to solve problems in research and industrial applications. • To learn the use of image processing and image understanding software tools. 14
Readings • Textbook: R. Klette. Concise Computer Vision: An Introduction into Theory and Algorithms, 1a Edition, Springer, 2014. • Textbook Website: http://ccv.wordpress.fos.auckland.ac.nz/ 15
Readings • About the Textbook: • It provides an introduction into basics of Computer vision and is not a guide on current research. • For a web>based introduction into topics in CV: http://homepages.inf.ed.ac.uk/rbf/CVonline/ 16
Readings • Complementary Books: R. Szeliski. Computer Vision: Algorithms and Applications, 1a Ed., Springer, 2010. R. Hartley and A. Zisserman. Multiple View Geometry in Computer Vision, 1a Ed., Cambridge Univ. Press, 2004. D. A. Forsyth and J. Ponce. Computer Vision: A Modern Approach, 1a Ed., Prentice Hall, 2002. 17
Assessment'and'Grading • Students will be assessed by using: • Problem Sets (Homework Assignments): 60%. • Final Project: 40%. • There are no exams. 18
Assignments • The assignments are comprised of problem sets (PS) and a final project (FP). • See the schedule for more details. • Assignments are posted online and are due in class by the end of the specified day of the lecture. All assignments must be handed in. 19
Assignments • They are designed to give you both theoretical and practical experience with the material. • Parts of the assignments will involve programming and experimentation. • It is allowed to use any coding environment that is convenient for you. 20
Final&Project • The final project allow you to explore a topic covered in class, which you found interesting. • The topic for the final project and its scope should be approved by myself previously. • It is comprised of (a) a project proposal , (b) a class presentation , and (c) a repor t. 21
Final&Project • Project Proposal: • The proposals should be just a page, and should describe what you plan to do. • In the proposal, lay out the tasks and give a timeline for when you will do each task. • You can work by yourself or in pairs. Projects by pairs should be correspondingly more substantial. 22
Final&Project • Project Presentation: • It should be clear, informative, and short. • You should briefly describe the problem and present an overview of your approach and results. • The time allotted to each presentation is 10 minutes. • Send the presentation to cardeal@cefetmg.br by the due time established. 23
Final&Project • Project Report: • The report should be 5 4 8 pages in SIBGRAPI format. • It should be structured like a research paper, with sections for introduction, related work, the approach, experimental results, conclusions and references. • Submit your report to cardeal@decom.cefetmg.br as a pdf file named YOUR_LAST_NAME.pdf. 24
Late%Policy • You have up to 5 late days for all assignments and you can use them at your discretion. • Any additional unapproved late submission will be considered as unsubmitted work. • Late submission is not allowed for the final project’s proposal. 25
Collaboration*Policy • I allow discussing problem sets with one or two classmates, but you must submit your own write7up and list your collaborators. • You are allowed to collaborate with one more student for the final project. 26
Course'Schedule • See the schedule for more details in: https://goo.gl/JT5CRR 27
What%is%Computer%Vision? • It studies how to reconstruct and understand a 3D scene from its 2D images, considering the features of structures present in the scene. • What kind of features ? 28
What%kind%of%features? • Geometric Features . Points, curves, surfaces and volumes, allowing to estimate, for example, shapes and positions of objects. p • Dynamic Features v . For example: velocity and acceleration. s 29
Semantic)Gap • One important goal of Computer Vision is to bridge the gap between pixels and “meaning”. What'a'computer'sees What'we'see 30 Source:'S.'Narasimhan
Why$is$it$so$hard? • What is in this image? - A hand holding a man? - A hand holding a mirrored sphere? - An Escher drawing? • Interpretations are ambiguous - The forward problem (graphics) is well-posed. - The “inverse problem” (vision) is not. 31 Source:*J.*Britton
What%is%it%related%to? Cognitive/ Neuroscience Sciences Computer/ Robotics Graphics Computer/ Vision Image/ Information/ Processing Retrieval Speech/ Machine/ Processing Learning 32
What%is%it%related%to? • Diagram(illustrating(the(work(scopes(of(Computer(Vision,( Computer(Graphics(and(Image(Processing. Image Computer#Graphics Processing 3D#World# Images Computer#Vision 33 33
Main%Components + + Cameras Computational Platform Software 34
Computer)X)Human)Vision Sensing(Device Interpreting(Device Interpretations Image((or(video) people,(line, grass,(tree,( buildings,( spring,(etc. 35
Debate&Moment • Should Computer Vision follow from our understanding of Human Vision? 8 Human9vision9“works”,9and9copying9is9easier9than9creatingA Yes 8 In9trying9to9mimic9human9vision,9we9learn9about9it. 8 There9are9several9different9biological vision systemsA 8 There9are9several9sensing9mechanismsA No 8 Synthetic9vision9systems9may9use9different9techniques9that999 are9more9appropriate9to9computational mechanisms. 36
Why$study$Computer$Vision? • Images and videos are everywhere. • Fast4growing collection of useful applications. • Several attractive scientific mysteries: • For instance: how does object recognition work? • Greater understanding of human vision. 37
Next%Lecture • Images in the Spatial Domain Pixels and Windows. Image Values and Basic Statistics. Spatial and Temporal Data Measures. Step>Edges. • Suggested reading Section 1.1 of textbook. 42
Recommend
More recommend
