Columbia University. Electrical Engineering Department. Fall 1999

Size: px
Start display at page:

Download "Columbia University. Electrical Engineering Department. Fall 1999"

Transcription

1 Columbia University Electrical Engineering Department Fall 1999 Report of the Project: Knowledge Based Semantic Segmentation Using Evolutionary Programming Professor: Shih-Fu Chang Student: Manuel J. Reyes.

2 Project: Knowledge-Based Semantic Image Segmentation Using Evolutionary Programming. Professor: Shih-Fu Chang. Student: Manuel Reyes. INTRODUCTION The Visual Apprentice is a system developed in the Image and Advanced TV Laboratory, which uses a Model-Based Classification of Visual Information for Content-Based Retrieval for still images and video.[1] Given the importance of objects and their parts in classification of visual information[2], the system bases its framework on an objectdefinition hierarchy of the following levels: (1) region; (2) perceptual; (3) object-part; (4) objects and (5) scene, as can be seen in figure 1[1]. During training the user selects the parameters of the segmentation, creates the class-definition hierarchy by defining the labels to be used from level 2 to level 5 and labels the regions in each training image/video frame according to the hierarchy. Level 4: Object Object Level 3: Object-part Object-Part1 Object-Part2 Object-Part3 Level 2: Perceptual Area Perceptual Area-1... Perceptual Area-n Level 1: Region Region-1... Region-n Figure 1: Every node in the tree has a conceptual meaning, but also corresponds to a set of connected pixels in the image. Using the information obtained during training, the system computes a feature vector for every element in the tree. These feature vectors are used to generate a set of fuzzy classifiers at different levels. As can be observed in figure 2 (scene with one object and two objectparts), classification follows the bottom up order of the object definition, the multiple level classifier structure is not identical to the object definition hierarchy; it has only one region classifier as a descendant to each object-part classifier. The first step in the classification of a new image/video is its automatic segmentation. This is done using the segmentation parameters provided by the user during the class definition and training phase. Once the input image/video is segmented, a region level classifier is applied, finding the grades of membership of each region to a particular object-part, so the regions that are more likely to be part of a particular object-part are found. Then adjoining groups of those regions are found. A perceptual area classifier is applied to that set of groups, which find the best combinations. Those combinations are in

3 turn combined to form prospective object-part, which form objects that form scenes.[1] Object-Part 1 Region Classifier Object-Part 2 Region Classifier Perceptual area 1 Perceptual area 2 Perceptual area 3 Perceptual area 1 Perceptual area 2 Object-Part 1 Object-Part 1 Object Scene Figure 2: Multiple Level classifiers models. Each classifier acts as a black box passing its results to the next level, thus allowing the incorporation of different learning and grouping strategies. Each box represents a classifier and arrows indicate the propagation of classification information.[1] The region classifier assigns a degree of membership to each region with respect to a particular object-part. Among the regions with membership value different than zero, a classification between strong(s) and weak(w) members is made (figure 3). Groups of these regions should be made to find the prospective perceptual areas. The search over the space of possible groups is done using Evolution Algorithms where the strong and weak candidates form the initial population. a) b) Figure 3: a) Segmented input image, b) Classified regions.

4 INITIAL PROPOSAL The objective of the project is to improve the search of the best groups of regions to find the prospective perceptual areas and to extend this improvement to the search of Object-Parts and Objects. The first step in the classification of a new image/video is its automatic segmentation; this is done using the segmentation parameters provided by the user during the training phase. Unfortunately, these parameters are not always the best ones for the new image/video and we could find regions that could be part of more than one object or that cover part of the object but also part of some other things not define in the object-definition hierarchy. An extreme case can be seen in the Figures 4 and 5. In figure 5, we can observe the automatic segmentation results of figure 4; there we can see that, there is a region that is part of the face, the hair and the hand of the woman. Figure 4: Original Image This part of the region corresponds to the hair of the women. This part of the region corresponds to the face of the women. This part of the region corresponds to the hand of the women. Figure 5: Segmented Image

5 In order to alleviate this problem, I suggested to subdivide the regions obtained from the automatic segmentation in to sub regions and then find the best groups of sub regions to obtain potential perceptual areas. The subdivision of the regions should be done using the feature vector calculated by the system for each one of the regions. The search of these groups was proposed to be done using evolution programming. We would use the segmented image in figure 3 to explain the proposed procedure. After classification the strong and weak regions would be numerated starting by number 1, and then subdivided in a certain number of sub regions (The subdivision processes implemented would be discuss later). These processes could be observed in Figure 6. a) b) Figure 6: a) Strong and weak regions numerated; b) Regions subdivided in sub regions, note that each region is divided in different patterns and in different number of sub regions. PROJECT DEVELOPMENT, IDEAS, IMPLEMENTATION AND RESULTS. The existing code to make the automatic segmentation[3] and to compute the feature vector for each region was first carefully studied. The understanding of how the regions are represented by the algorithm is crucial for the development of this project. In order to become more familiar with the representation of regions, I decided to work first in the search of group of regions, because each group is another region itself and it should be represented in the same form as in the case of a single region. The philosophy of the search algorithm is the same, if it is use with regions or with sub regions, so I decided to implement the search algorithm with regions first, later I will explain the subdivision processes implemented. Although, originally, I proposed to use evolution programming to realize the search, I decided first to work with a traditionally search algorithm such as Greedy Search. Because of it would be easier to manage the region representation in this algorithm than in an evolution one.

6 Greedy Search Algorithm In the greedy search algorithm, the following terms are used, state: defines a particularly situation of the system; initial state: initial situation in the system; goal state: the desired situation; operator: is used to denote the description of an action in terms of which state will be reached by carrying out the action in a particular state, when it is applied to a state, it generates a different set of states called successors of the state, this process is known as the expansion of a state; and finally an heuristic function, which provides a measure of how close is a particularly state to the goal one[4]. The algorithm starts with the initial state, then the initial state is expanded to get a new set of states, the heuristic function is applied to the whole state space to determine which is the closer state to the goal one, then that state is expanded, the algorithm determines again which is now the best state and expands it, and so on. The algorithm stops when it reaches a state with a particular heuristic value (goal state test), after a predefined number of expanded states or when there is not any improvement in the heuristic function in the newest successors[4]. It is important to notice that the program should keep in memory the whole state space to determine which state is the best one, it also has to keep track of which states have been already expanded and be careful to avoid repetition of states. In this project, a state is a particular group of regions (sub regions), the initial state is formed by all the regions classified as strong or weak. The operator will eliminate one of the regions of the state, so when a state with n regions is expanded, n different successors with n-1 regions are generated. The heuristic function utilized in this case is the perceptual area classifier A graphic chart of the algorithm for the segmented image of figure 6 b) can be observed in figure 7. The heuristic function used in this example is the roundness of the region. In the segmentation algorithm used the extension of a region is represented by a mask, which occupied a lot of memory. If the regions represented by every state were represented in the same way, the amount of memory needed to keep the whole state space would be enormous. So I represent the states using an id number, which is formed concatenating the number of each one of the original regions that the state contains, in the figure 7 some of these id numbers can be observed. The mask of the region will be computed only when the heuristic function is going to be applied to the corresponding state, but the program doesn t keep the mask in memory. Using the id number, I avoid the repetition of states. As we can see in figure 7, the algorithm would find the best group of regions (Id: ) for this particular heuristic function, but it would also expand it, after this expansion the algorithm would not be able to find another state with a highest heuristic value than the one found for state So, I used the following criteria, if after 5 consecutive state expansions the best value of the heuristic value doesn t improve, the algorithm would stop.

7 Initial state 1 9 Id: First Expansion.... Id: Id: Id: Id: Second Expansion Fifth Expansion.... Id: Id: Third Expansion.. Id: Id: Fourth Expansion.. Id: Id: Figure 7: Example of the greedy search algorithm implemented, the order of the state expansions is determined by the heuristic function, the id number of the states represents the original regions that the state contains, this id number is used to avoid the repletion of states. The evolution program would use the same representation of states than in the greedy search algorithm, but instead of expanding states, it would mutate them. Although theoretically this implies a big difference, I think that in the practice it is not that much, so I decided to work in the subdivision of the regions rather than in the implementation of the evolution-programming search. Subdivision of Regions The subdivision of regions is made in order to improve the results of the automatic segmentation applied to any new input image/video. Analyzing different segmented images; I found basically three types of segmentation errors as can be seen in figure 8. Type I, all the regions of the object-part cover other portions of the image; type II, some regions are totally enclosed by the object, but others cover also other portions of the image; type III, a single region covers the object-part and also other portion of the image.

8 Object-Part a) b) c) Figure 8: Segmentation errors a) Type I, b) Type II c) Type III. I decided to work first with the third case; an example of this case can be seen in figure 9. The region that corresponds to the face covers also part of the background of the image. a) b) Figure 9: a) Original image, b) Segmented image. First Approach The image segmentation process creates three feature maps from the original image: edge map, color map and motion field. Color map is the major feature map, it is generated by first converting the original image into the CIE L*u*v* color space and then quantizing pixels to a limited number of colors using a clustering based method. The edge map is a binary mask where edge pixels are set to 1 and non-edge-pixels are set to 0. It is generated by applying the Canny edge detection algorithm. The segmentation algorithm fusions color and edge information to improve the segmentation process[]. As I mentioned before in this case the region covers the object and also other portions of the image. Part of the problem is that the object part and the other portions of the image covered by the region

9 are merged into a single quantization level during the generation of the color map. My first approach consists in reapplied the segmentation process only to the bounding box of the selected region. Because of I am using only the local information, the quantization levels in the generation of the color map would change and then the selected region would be re-segmented. Then the new regions obtained in the position of the selected region would be considered as the sub regions. This approach would not work if the distribution of the colors in the bounding box were similar to the distribution in the whole image. To prevent this situation, I developed two alternatives; the bounding box contains the selected region and part of the background, so if I change the original background to black or white, the distribution of color in the bounding box would change significantly. In figure 10, we can observed the images containing the bounding box of the face region of figure 9 using a) the original background; b) white background and c) black background. a) b) c) Figure 10: Image containing the bounding box of the face region of figure 9 using a) the original background, b) white background and c) black background. Once the image with the bounding box is segmented, the masks of the new regions are mapped to the mask of the original selected region, having then the desired subdivision. The results of the subdivision of the selected region in the whole segmented image, for the three cases can be observed in figure 11. a) b) c) Figure 11: Mapping in the entire segmented image, the results of the resegmentation of the image of the bounding box using a) original background, b) white background and c) black background.

10 Another two examples of this approach can be observed in figures 12 and 13. a) Original image b) Segmented image c) d)

11 e) Figure 12: Mapping in the entire segmented image, the results of the resegmentation of the image of the bounding box using c) original background, d) black background and e) white background. a) Original Image b)segmented image c)

12 d) e) Figure 13: Mapping in the entire segmented image, the results of the resegmentation of the image of the bounding box using c) original background, d) black background and e) white background. As we can see, the results using different backgrounds differ for each example. I thought in two forms for the automation of the background s choice: One is to estimate the color distribution of the input image, then calculate the color distribution of the image of the bounding box using the three different backgrounds and then use the background for which the color distribution would be more different than the color distribution of the input image. The other form was to observe the color features and the information about which background produced a better re-segmentation for each image in a training set (I used 30 images) and compute a decision tree to decide which background to use for a new image using its color features as the decision factor. But after discussing these ideas with Alex Jaimes, we agreed that for both cases the automation would be very dependant of the segmentation algorithm and that this would subtract flexibility to the system, so I decided to work in another more general approach for the subdivision of the regions. Comments about the first approach. Even though, the first approach I used for the subdivision resolves only a particular type of error and that it is dependant of the segmentation algorithm I thought in some applications were it could be used. Suppose that we want to train the system to recognize a set of objects that normally are grouped together in the image and that there are also some other objects in the image. If we want to segment every one of the objects of our interest we would need to choose the parameters of the segmentation in the training phase such that the segmentation would be accurate enough to segment all the objects in the image, this would produce a lot of regions including the ones of the objects in which we are not interested on. These undesired regions are going to be classified consuming time during the classification phase. Now if we train the system using a coarse segmentation such that the group of objects of our interest would be included in a single region, the total number of regions in the segmented image would be reduced. If the

13 system is trained to recognized the region that include the group of objects of our interest, a local segmentation could be use in that region to segment the region in the different objects it contains and then apply a second classification for the regions of the desired objects. For example in figure 14, we may want to recognize the batter, the catcher and the umpire, if we use an accurate segmentation, the pitcher, his body parts and some persons in the public could be segmented as well, having in this case to many regions to classify, but if we tune the system with the segmentation parameters that would segment the group of desired objects as a single one, the resulting number of regions would be considerable small. Then, the local segmentation could be performed using the same parameters or using a finer segmentation. Objects of interest Figure 14 The second idea developed in this project for the subdivision of the regions is discuss in the next pages.

14 Second Approach. After observing hundred of images a second idea came along. Frequently in segmented images two or more different parts of the image are united into a single region by a few number of pixels. A critical example can be seen in figure 15. a) b) Figure 15: a) Original Image, b) Segmented image: the regions of the faces are united by a relative small number of pixels. One of the features that the present system calculates is the maximum line (or distance) of connected pixels in the region. As it can be observed in Figure 16, the minimum is estimated using the minor axis of the ellipse that best fits the region in the orientation of the maximum line. Maximum line Figure 16: Maximum line of the region of interest of figure 15 b)

15 There is a function in the present system that calculates the largest line of the region in the horizontal direction. In order to find the maximum line the region is rotated using 16 different angles ( θ = 22.5 ), the largest horizontal line is calculated for each rotation and the largest one is the maximum line (For now on, I would use the term distance to mean a line of connected pixels in the horizontal direction). Now, what I want is to find the points of the region were a relative small number of pixels unite comparative bigger portions of the region and then divide the region in those points. This means that the distance of the region before and after those points is larger than the distance in those points, then those points would correspond to a relative minimum of the function of the region distance. (The graphics of the distance function for the region of figure 16 for different rotation angles can be observed in the appendix A of this report), an example for the region of figure 16 rotated by an angle of 90 can be observed in figure 17. Another relative minimum The distance after point X is bigger than the one in point X The distance before point X is bigger than the one in point X X, Relative minimum of the distance Figure 17: Region of figure 17 rotated by an angle of 90, the point X corresponds to a relative minimum of the distance for this region. More than relative minimum could be found for a particular rotated region. To find the relative minimum of the function distance of each rotated region, I compute the distance function using the existing function that calculates connected horizontal lines in the region, then I find its derivative and using the first derivative criteria I find the relative minimums. From this relative minimums, I find the absolute minimum and this point is marked as a potential point of division. The original function uses 16 different angles ( θ = 22.5 ), completing a rotation interval from 0 to 360, but to calculate the minimum we only need 8 different rotation angles ( θ = 22.5 ) completing a rotation interval from 0 to 180, because the distance functions in the interval from 180 to 360 are exactly the reflection in time of the ones in the first interval and then, they have exactly the same minimums, as it can be observed in the graphics of the appendix A. For some angles of rotation, the absolute minimum of the region distance is not significantly smaller than the largest distance found for the same rotated region, in those cases the absolute minimum would not be used as a potential point of division. For this project the discrimination criteria is the ratio between the absolute minimum and the largest distance, if it is not smaller than 0.5, that particular absolute minimum is eliminated as a potential point of division.

16 The suggested subdivision for the region in figure 19 using the significant absolute minimums found for all the rotations can be seen in figure 18. As we can see a better division could be done if I would permit more than one point of division per rotation. This is one of the open issues in this approach. This point would be also marked if I would use more than one point for each rotation Another example can be observed in figure 20. a) b) c) Figure 20: a) original image b) Region in which the program was used c) proposed division of the region. This approach can be used only in regions were significant minimum of the distance functions exists, for example for a round region this

17 approach cannot be used. Even though, I presented two examples of the third kind of errors, this approach could be used in any region as long as a significant minimum of the distance exists. These values (The minimums) would be computed as a part of the feature vector of the regions. Comments about the second approach For this approach, I only find the potential points of division, but I didn t have time to divide the region and to generate the masks of the subsequent sub regions. As I commented before, one of the open issues for this approach is the number of minimums to use for each rotation, I have only work with the absolute minimum but, it could be extended to work with more than one. Another detail is the discrimination factor for significant minimums. One detail I have not mentioned before is that, frequently a region has more than one connected line for a given horizontal position, when I found this case, I decided to use the largest one for the computation of the minimums, but different calculation could be done increasing the performance of the region division. Final Comments. The greedy search algorithm developed in the first part of the project can be used with the sub regions that can be obtained using the first approach to divide the regions, this is not the case for the second approach where some work has to be done to apply the search algorithm to the sub regions that may be obtained from this approach. I think that the main issue in this project is the division of the regions; this is a very challenging and not trivial problem with a lot of variants. I think that, it is an open problem where a lot of different ideas could be implemented. References [1] A. Jaimes and S.-F. Chang, Model-Based Classification of Visual Information for Content-Based Retrieval, Storage and Retrieval for Image and Video Databases VII, IS&T/SPIE99, San Jose, CA. January 1999 [2] A. Jaimes and S.-F. Chang, Syntax, Semantics and Visual Information: A Conceptual Representation for Indexing, ADVENT project technical report No , Columbia University, January [3] D. Zhong and S.-F. Chang, Video Object Model and Segmentation for content Based Video Indexing, 1997 IEEE International Symposium of Circuits and Systems, June [4] Stuart Russell and Peter Norvig, Artificial Intelligence, A Modern Approach, Prentice Hall, 1995.

18 APPENDIX A

Clustering Methods for Video Browsing and Annotation

Clustering Methods for Video Browsing and Annotation Clustering Methods for Video Browsing and Annotation Di Zhong, HongJiang Zhang 2 and Shih-Fu Chang* Institute of System Science, National University of Singapore Kent Ridge, Singapore 05 *Center for Telecommunication

More information

Real-Time Content-Based Adaptive Streaming of Sports Videos

Real-Time Content-Based Adaptive Streaming of Sports Videos Real-Time Content-Based Adaptive Streaming of Sports Videos Shih-Fu Chang, Di Zhong, and Raj Kumar Digital Video and Multimedia Group ADVENT University/Industry Consortium Columbia University December

More information

Final Review CMSC 733 Fall 2014

Final Review CMSC 733 Fall 2014 Final Review CMSC 733 Fall 2014 We have covered a lot of material in this course. One way to organize this material is around a set of key equations and algorithms. You should be familiar with all of these,

More information

ELEN E4830 Digital Image Processing. Homework 6 Solution

ELEN E4830 Digital Image Processing. Homework 6 Solution ELEN E4830 Digital Image Processing Homework 6 Solution Chuxiang Li cxli@ee.columbia.edu Department of Electrical Engineering, Columbia University April 10, 2006 1 Edge Detection 1.1 Sobel Operator The

More information

Cs : Computer Vision Final Project Report

Cs : Computer Vision Final Project Report Cs 600.461: Computer Vision Final Project Report Giancarlo Troni gtroni@jhu.edu Raphael Sznitman sznitman@jhu.edu Abstract Given a Youtube video of a busy street intersection, our task is to detect, track,

More information

Compilers and Interpreters

Compilers and Interpreters Overview Roadmap Language Translators: Interpreters & Compilers Context of a compiler Phases of a compiler Compiler Construction tools Terminology How related to other CS Goals of a good compiler 1 Compilers

More information

FPGA Implementation of a Nonlinear Two Dimensional Fuzzy Filter

FPGA Implementation of a Nonlinear Two Dimensional Fuzzy Filter Justin G. R. Delva, Ali M. Reza, and Robert D. Turney + + CORE Solutions Group, Xilinx San Jose, CA 9514-3450, USA Department of Electrical Engineering and Computer Science, UWM Milwaukee, Wisconsin 5301-0784,

More information

NOTES ON OBJECT-ORIENTED MODELING AND DESIGN

NOTES ON OBJECT-ORIENTED MODELING AND DESIGN NOTES ON OBJECT-ORIENTED MODELING AND DESIGN Stephen W. Clyde Brigham Young University Provo, UT 86402 Abstract: A review of the Object Modeling Technique (OMT) is presented. OMT is an object-oriented

More information

Efficient memory-bounded search methods

Efficient memory-bounded search methods Efficient memory-bounded search methods Mikhail Simin Arjang Fahim CSCE 580: Artificial Intelligence Fall 2011 Dr. Marco Voltorta Outline of The Presentation Motivations and Objectives Background - BFS

More information

Identifying and Reading Visual Code Markers

Identifying and Reading Visual Code Markers O. Feinstein, EE368 Digital Image Processing Final Report 1 Identifying and Reading Visual Code Markers Oren Feinstein, Electrical Engineering Department, Stanford University Abstract A visual code marker

More information

PARAMETER OPTIMIZATION FOR AUTOMATED SIGNAL ANALYSIS FOR CONDITION MONITORING OF AIRCRAFT SYSTEMS. Mike Gerdes 1, Dieter Scholz 1

PARAMETER OPTIMIZATION FOR AUTOMATED SIGNAL ANALYSIS FOR CONDITION MONITORING OF AIRCRAFT SYSTEMS. Mike Gerdes 1, Dieter Scholz 1 AST 2011 Workshop on Aviation System Technology PARAMETER OPTIMIZATION FOR AUTOMATED SIGNAL ANALYSIS FOR CONDITION MONITORING OF AIRCRAFT SYSTEMS Mike Gerdes 1, Dieter Scholz 1 1 Aero - Aircraft Design

More information

Visualization of Crowd-Powered Impression Evaluation Results

Visualization of Crowd-Powered Impression Evaluation Results Visualization of Crowd-Powered Impression Evaluation Results Erika GOMI,YuriSAITO, Takayuki ITOH (*)Graduate School of Humanities and Sciences, Ochanomizu University Tokyo, Japan {erika53, yuri, itot }

More information

Cross Reference Strategies for Cooperative Modalities

Cross Reference Strategies for Cooperative Modalities Cross Reference Strategies for Cooperative Modalities D.SRIKAR*1 CH.S.V.V.S.N.MURTHY*2 Department of Computer Science and Engineering, Sri Sai Aditya institute of Science and Technology Department of Information

More information

6. Applications - Text recognition in videos - Semantic video analysis

6. Applications - Text recognition in videos - Semantic video analysis 6. Applications - Text recognition in videos - Semantic video analysis Stephan Kopf 1 Motivation Goal: Segmentation and classification of characters Only few significant features are visible in these simple

More information

Segmentation algorithm for monochrome images generally are based on one of two basic properties of gray level values: discontinuity and similarity.

Segmentation algorithm for monochrome images generally are based on one of two basic properties of gray level values: discontinuity and similarity. Chapter - 3 : IMAGE SEGMENTATION Segmentation subdivides an image into its constituent s parts or objects. The level to which this subdivision is carried depends on the problem being solved. That means

More information

Ghassan Samara Internet Technology Department Zarqa University, Jordan.

Ghassan Samara Internet Technology Department Zarqa University, Jordan. World of Computer Science and Information Technology Journal (WCSIT) ISSN: 2221-0741 Vol. 7, No. 2, 10-19, 2017 A Practical Approach for Detecting Logical Error in Object Oriented Environment Ghassan Samara

More information

Subset sum problem and dynamic programming

Subset sum problem and dynamic programming Lecture Notes: Dynamic programming We will discuss the subset sum problem (introduced last time), and introduce the main idea of dynamic programming. We illustrate it further using a variant of the so-called

More information

Introduction to the Practice of Statistics Fifth Edition Moore, McCabe

Introduction to the Practice of Statistics Fifth Edition Moore, McCabe Introduction to the Practice of Statistics Fifth Edition Moore, McCabe Section 1.3 Homework Answers Assignment 5 1.80 If you ask a computer to generate "random numbers between 0 and 1, you uniform will

More information

(Refer Slide Time: 00:01:30)

(Refer Slide Time: 00:01:30) Digital Circuits and Systems Prof. S. Srinivasan Department of Electrical Engineering Indian Institute of Technology, Madras Lecture - 32 Design using Programmable Logic Devices (Refer Slide Time: 00:01:30)

More information

MAT128A: Numerical Analysis Lecture Two: Finite Precision Arithmetic

MAT128A: Numerical Analysis Lecture Two: Finite Precision Arithmetic MAT128A: Numerical Analysis Lecture Two: Finite Precision Arithmetic September 28, 2018 Lecture 1 September 28, 2018 1 / 25 Floating point arithmetic Computers use finite strings of binary digits to represent

More information

Artificial Intelligence

Artificial Intelligence Torralba and Wahlster Artificial Intelligence Chapter 8: Constraint Satisfaction Problems, Part I 1/48 Artificial Intelligence 8. CSP, Part I: Basics, and Naïve Search What to Do When Your Problem is to

More information

ORGANIZING THE DATA IN A FREQUENCY TABLE

ORGANIZING THE DATA IN A FREQUENCY TABLE ORGANIZING THE DATA IN A FREQUENCY TABLE Suppose the scores obtained by 5 students on a standardized test are as follows: 68, 55, 61, 55, 43, 59, 55, 58, 77, 6, 56, 53, 58, 7, 57, 62, 5, 69, 44, 63, 48,79,

More information

FACILITATING INFRARED SEEKER PERFORMANCE TRADE STUDIES USING DESIGN SHEET

FACILITATING INFRARED SEEKER PERFORMANCE TRADE STUDIES USING DESIGN SHEET FACILITATING INFRARED SEEKER PERFORMANCE TRADE STUDIES USING DESIGN SHEET Sudhakar Y. Reddy and Kenneth W. Fertig Rockwell Science Center, Palo Alto Laboratory Palo Alto, California and Anne Hemingway

More information

Color Image Segmentation

Color Image Segmentation Color Image Segmentation Yining Deng, B. S. Manjunath and Hyundoo Shin* Department of Electrical and Computer Engineering University of California, Santa Barbara, CA 93106-9560 *Samsung Electronics Inc.

More information

Motion Detection Algorithm

Motion Detection Algorithm Volume 1, No. 12, February 2013 ISSN 2278-1080 The International Journal of Computer Science & Applications (TIJCSA) RESEARCH PAPER Available Online at http://www.journalofcomputerscience.com/ Motion Detection

More information

TIMSS 2011 Fourth Grade Mathematics Item Descriptions developed during the TIMSS 2011 Benchmarking

TIMSS 2011 Fourth Grade Mathematics Item Descriptions developed during the TIMSS 2011 Benchmarking TIMSS 2011 Fourth Grade Mathematics Item Descriptions developed during the TIMSS 2011 Benchmarking Items at Low International Benchmark (400) M01_05 M05_01 M07_04 M08_01 M09_01 M13_01 Solves a word problem

More information

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm Group 1: Mina A. Makar Stanford University mamakar@stanford.edu Abstract In this report, we investigate the application of the Scale-Invariant

More information

Improving the Efficiency of Fast Using Semantic Similarity Algorithm

Improving the Efficiency of Fast Using Semantic Similarity Algorithm International Journal of Scientific and Research Publications, Volume 4, Issue 1, January 2014 1 Improving the Efficiency of Fast Using Semantic Similarity Algorithm D.KARTHIKA 1, S. DIVAKAR 2 Final year

More information

Layout Segmentation of Scanned Newspaper Documents

Layout Segmentation of Scanned Newspaper Documents , pp-05-10 Layout Segmentation of Scanned Newspaper Documents A.Bandyopadhyay, A. Ganguly and U.Pal CVPR Unit, Indian Statistical Institute 203 B T Road, Kolkata, India. Abstract: Layout segmentation algorithms

More information

6. Dicretization methods 6.1 The purpose of discretization

6. Dicretization methods 6.1 The purpose of discretization 6. Dicretization methods 6.1 The purpose of discretization Often data are given in the form of continuous values. If their number is huge, model building for such data can be difficult. Moreover, many

More information

A New Algorithm for Shape Detection

A New Algorithm for Shape Detection IOSR Journal of Computer Engineering (IOSR-JCE) e-issn: 2278-0661,p-ISSN: 2278-8727, Volume 19, Issue 3, Ver. I (May.-June. 2017), PP 71-76 www.iosrjournals.org A New Algorithm for Shape Detection Hewa

More information

Practical Image and Video Processing Using MATLAB

Practical Image and Video Processing Using MATLAB Practical Image and Video Processing Using MATLAB Chapter 18 Feature extraction and representation What will we learn? What is feature extraction and why is it a critical step in most computer vision and

More information

Clustering Color/Intensity. Group together pixels of similar color/intensity.

Clustering Color/Intensity. Group together pixels of similar color/intensity. Clustering Color/Intensity Group together pixels of similar color/intensity. Agglomerative Clustering Cluster = connected pixels with similar color. Optimal decomposition may be hard. For example, find

More information

THE REAL NUMBER SYSTEM

THE REAL NUMBER SYSTEM THE REAL NUMBER SYSTEM Review The real number system is a system that has been developing since the beginning of time. By now you should be very familiar with the following number sets : Natural or counting

More information

Ray Tracing Acceleration Data Structures

Ray Tracing Acceleration Data Structures Ray Tracing Acceleration Data Structures Sumair Ahmed October 29, 2009 Ray Tracing is very time-consuming because of the ray-object intersection calculations. With the brute force method, each ray has

More information

Edge detection. Stefano Ferrari. Università degli Studi di Milano Elaborazione delle immagini (Image processing I)

Edge detection. Stefano Ferrari. Università degli Studi di Milano Elaborazione delle immagini (Image processing I) Edge detection Stefano Ferrari Università degli Studi di Milano stefano.ferrari@unimi.it Elaborazione delle immagini (Image processing I) academic year 2011 2012 Image segmentation Several image processing

More information

Dynamic Obstacle Detection Based on Background Compensation in Robot s Movement Space

Dynamic Obstacle Detection Based on Background Compensation in Robot s Movement Space MATEC Web of Conferences 95 83 (7) DOI:.5/ matecconf/79583 ICMME 6 Dynamic Obstacle Detection Based on Background Compensation in Robot s Movement Space Tao Ni Qidong Li Le Sun and Lingtao Huang School

More information

Eye Localization Using Color Information. Amit Chilgunde

Eye Localization Using Color Information. Amit Chilgunde Eye Localization Using Color Information Amit Chilgunde Department of Electrical and Computer Engineering National University of Singapore, Singapore ABSTRACT In this project, we propose localizing the

More information

Chapter 2: Understanding Data Distributions with Tables and Graphs

Chapter 2: Understanding Data Distributions with Tables and Graphs Test Bank Chapter 2: Understanding Data with Tables and Graphs Multiple Choice 1. Which of the following would best depict nominal level data? a. pie chart b. line graph c. histogram d. polygon Ans: A

More information

APPLICATION OF AERIAL VIDEO FOR TRAFFIC FLOW MONITORING AND MANAGEMENT

APPLICATION OF AERIAL VIDEO FOR TRAFFIC FLOW MONITORING AND MANAGEMENT Pitu Mirchandani, Professor, Department of Systems and Industrial Engineering Mark Hickman, Assistant Professor, Department of Civil Engineering Alejandro Angel, Graduate Researcher Dinesh Chandnani, Graduate

More information

Image Mosaicing with Motion Segmentation from Video

Image Mosaicing with Motion Segmentation from Video Image Mosaicing with Motion Segmentation from Video Augusto Román and Taly Gilat EE392J Digital Video Processing Winter 2002 Introduction: Many digital cameras these days include the capability to record

More information

Color quantization using modified median cut

Color quantization using modified median cut Color quantization using modified median cut Dan S. Bloomberg Leptonica Abstract We describe some observations on the practical implementation of the median cut color quantization algorithm, suitably modified

More information

EDIT202 Spreadsheet Lab Prep Sheet

EDIT202 Spreadsheet Lab Prep Sheet EDIT202 Spreadsheet Lab Prep Sheet While it is clear to see how a spreadsheet may be used in a classroom to aid a teacher in marking (as your lab will clearly indicate), it should be noted that spreadsheets

More information

Periodic Pattern Detection for Real-Time Application

Periodic Pattern Detection for Real-Time Application Periodic Pattern Detection for Real-Time Application Giovanni Puglisi 1 and Sebastiano Battiato 1 Dipartimento di Matematica e Informatica University of Catania, Italy {puglisi,battiato}@dmi.unict.it Abstract.

More information

Object Extraction Using Image Segmentation and Adaptive Constraint Propagation

Object Extraction Using Image Segmentation and Adaptive Constraint Propagation Object Extraction Using Image Segmentation and Adaptive Constraint Propagation 1 Rajeshwary Patel, 2 Swarndeep Saket 1 Student, 2 Assistant Professor 1 2 Department of Computer Engineering, 1 2 L. J. Institutes

More information

GAP CLOSING. Grade 9. Facilitator s Guide

GAP CLOSING. Grade 9. Facilitator s Guide GAP CLOSING Grade 9 Facilitator s Guide Topic 3 Integers Diagnostic...5 Administer the diagnostic...5 Using diagnostic results to personalize interventions solutions... 5 Using Intervention Materials...8

More information

Research on QR Code Image Pre-processing Algorithm under Complex Background

Research on QR Code Image Pre-processing Algorithm under Complex Background Scientific Journal of Information Engineering May 207, Volume 7, Issue, PP.-7 Research on QR Code Image Pre-processing Algorithm under Complex Background Lei Liu, Lin-li Zhou, Huifang Bao. Institute of

More information

Lecture 10: Semantic Segmentation and Clustering

Lecture 10: Semantic Segmentation and Clustering Lecture 10: Semantic Segmentation and Clustering Vineet Kosaraju, Davy Ragland, Adrien Truong, Effie Nehoran, Maneekwan Toyungyernsub Department of Computer Science Stanford University Stanford, CA 94305

More information

CS 231A Computer Vision (Fall 2011) Problem Set 4

CS 231A Computer Vision (Fall 2011) Problem Set 4 CS 231A Computer Vision (Fall 2011) Problem Set 4 Due: Nov. 30 th, 2011 (9:30am) 1 Part-based models for Object Recognition (50 points) One approach to object recognition is to use a deformable part-based

More information

Basically, a graph is a representation of the relationship between two or more variables.

Basically, a graph is a representation of the relationship between two or more variables. 1 Drawing Graphs Introduction In recent years, the CSEC Integrated Science Examination, Paper 02, deals with graphical analysis. That is, data is presented in a tabular format and the student is asked

More information

Auto-Digitizer for Fast Graph-to-Data Conversion

Auto-Digitizer for Fast Graph-to-Data Conversion Auto-Digitizer for Fast Graph-to-Data Conversion EE 368 Final Project Report, Winter 2018 Deepti Sanjay Mahajan dmahaj@stanford.edu Sarah Pao Radzihovsky sradzi13@stanford.edu Ching-Hua (Fiona) Wang chwang9@stanford.edu

More information

Chapter 6 Rational Numbers and Proportional Reasoning

Chapter 6 Rational Numbers and Proportional Reasoning Chapter 6 Rational Numbers and Proportional Reasoning Students should build their understanding of fractions as parts of a whole and as division. They will need to see and explore a variety of models of

More information

Terrain Rendering Research for Games. Jonathan Blow Bolt Action Software

Terrain Rendering Research for Games. Jonathan Blow Bolt Action Software Terrain Rendering Research for Games Jonathan Blow Bolt Action Software jon@bolt-action.com Lecture Agenda Introduction to the problem Survey of established algorithms Problems with established algorithms

More information

Comparison between Various Edge Detection Methods on Satellite Image

Comparison between Various Edge Detection Methods on Satellite Image Comparison between Various Edge Detection Methods on Satellite Image H.S. Bhadauria 1, Annapurna Singh 2, Anuj Kumar 3 Govind Ballabh Pant Engineering College ( Pauri garhwal),computer Science and Engineering

More information

Structural and Syntactic Pattern Recognition

Structural and Syntactic Pattern Recognition Structural and Syntactic Pattern Recognition Selim Aksoy Department of Computer Engineering Bilkent University saksoy@cs.bilkent.edu.tr CS 551, Fall 2017 CS 551, Fall 2017 c 2017, Selim Aksoy (Bilkent

More information

Malaysian License Plate Recognition Artificial Neural Networks and Evolu Computation. The original publication is availabl

Malaysian License Plate Recognition Artificial Neural Networks and Evolu Computation. The original publication is availabl JAIST Reposi https://dspace.j Title Malaysian License Plate Recognition Artificial Neural Networks and Evolu Computation Stephen, Karungaru; Fukumi, Author(s) Minoru; Norio Citation Issue Date 2005-11

More information

CHAPTER 4: MICROSOFT OFFICE: EXCEL 2010

CHAPTER 4: MICROSOFT OFFICE: EXCEL 2010 CHAPTER 4: MICROSOFT OFFICE: EXCEL 2010 Quick Summary A workbook an Excel document that stores data contains one or more pages called a worksheet. A worksheet or spreadsheet is stored in a workbook, and

More information

A Generalized Method to Solve Text-Based CAPTCHAs

A Generalized Method to Solve Text-Based CAPTCHAs A Generalized Method to Solve Text-Based CAPTCHAs Jason Ma, Bilal Badaoui, Emile Chamoun December 11, 2009 1 Abstract We present work in progress on the automated solving of text-based CAPTCHAs. Our method

More information

Excel 2010: Getting Started with Excel

Excel 2010: Getting Started with Excel Excel 2010: Getting Started with Excel Excel 2010 Getting Started with Excel Introduction Page 1 Excel is a spreadsheet program that allows you to store, organize, and analyze information. In this lesson,

More information

Hierarchical Clustering 4/5/17

Hierarchical Clustering 4/5/17 Hierarchical Clustering 4/5/17 Hypothesis Space Continuous inputs Output is a binary tree with data points as leaves. Useful for explaining the training data. Not useful for making new predictions. Direction

More information

Aston Hall s A-Z of mathematical terms

Aston Hall s A-Z of mathematical terms Aston Hall s A-Z of mathematical terms The following guide is a glossary of mathematical terms, covering the concepts children are taught in FS2, KS1 and KS2. This may be useful to clear up any homework

More information

Net Masks and IP Addresses

Net Masks and IP Addresses Net Masks and IP Addresses Professor Don Colton Brigham Young University Hawaii 1 Introduction IPv4 is currently the main addressing method on the Internet. Students who plan to use networking skills in

More information

Motion. 1 Introduction. 2 Optical Flow. Sohaib A Khan. 2.1 Brightness Constancy Equation

Motion. 1 Introduction. 2 Optical Flow. Sohaib A Khan. 2.1 Brightness Constancy Equation Motion Sohaib A Khan 1 Introduction So far, we have dealing with single images of a static scene taken by a fixed camera. Here we will deal with sequence of images taken at different time intervals. Motion

More information

Time Stamp Detection and Recognition in Video Frames

Time Stamp Detection and Recognition in Video Frames Time Stamp Detection and Recognition in Video Frames Nongluk Covavisaruch and Chetsada Saengpanit Department of Computer Engineering, Chulalongkorn University, Bangkok 10330, Thailand E-mail: nongluk.c@chula.ac.th

More information

Image Compression Using BPD with De Based Multi- Level Thresholding

Image Compression Using BPD with De Based Multi- Level Thresholding International Journal of Innovative Research in Electronics and Communications (IJIREC) Volume 1, Issue 3, June 2014, PP 38-42 ISSN 2349-4042 (Print) & ISSN 2349-4050 (Online) www.arcjournals.org Image

More information

Predicting Messaging Response Time in a Long Distance Relationship

Predicting Messaging Response Time in a Long Distance Relationship Predicting Messaging Response Time in a Long Distance Relationship Meng-Chen Shieh m3shieh@ucsd.edu I. Introduction The key to any successful relationship is communication, especially during times when

More information

9 POINTS TO A GOOD LINE GRAPH

9 POINTS TO A GOOD LINE GRAPH NAME: PD: DATE: 9 POINTS TO A GOOD LINE GRAPH - 2013 1. Independent Variable on the HORIZONTAL (X) AXIS RANGE DIVIDED BY SPACES and round up to nearest usable number to spread out across the paper. LABELED

More information

A Parallel Evolutionary Algorithm for Discovery of Decision Rules

A Parallel Evolutionary Algorithm for Discovery of Decision Rules A Parallel Evolutionary Algorithm for Discovery of Decision Rules Wojciech Kwedlo Faculty of Computer Science Technical University of Bia lystok Wiejska 45a, 15-351 Bia lystok, Poland wkwedlo@ii.pb.bialystok.pl

More information

Lesson 15. Student Outcomes. Lesson Notes. Classwork. Opening (2 minutes) Opening Exercise (3 minutes) (optional)

Lesson 15. Student Outcomes. Lesson Notes. Classwork. Opening (2 minutes) Opening Exercise (3 minutes) (optional) Lesson 5 Lesson 5: Piecewise Functions Student Outcomes Students examine the features of piecewise functions including the absolute value function and step functions. Students understand that the graph

More information

Unsupervised Learning and Clustering

Unsupervised Learning and Clustering Unsupervised Learning and Clustering Selim Aksoy Department of Computer Engineering Bilkent University saksoy@cs.bilkent.edu.tr CS 551, Spring 2009 CS 551, Spring 2009 c 2009, Selim Aksoy (Bilkent University)

More information

(Refer Slide Time: 00:02:02)

(Refer Slide Time: 00:02:02) Computer Graphics Prof. Sukhendu Das Dept. of Computer Science and Engineering Indian Institute of Technology, Madras Lecture - 20 Clipping: Lines and Polygons Hello and welcome everybody to the lecture

More information

Your Flowchart Secretary: Real-Time Hand-Written Flowchart Converter

Your Flowchart Secretary: Real-Time Hand-Written Flowchart Converter Your Flowchart Secretary: Real-Time Hand-Written Flowchart Converter Qian Yu, Rao Zhang, Tien-Ning Hsu, Zheng Lyu Department of Electrical Engineering { qiany, zhangrao, tiening, zhenglyu} @stanford.edu

More information

9.913 Pattern Recognition for Vision. Class 8-2 An Application of Clustering. Bernd Heisele

9.913 Pattern Recognition for Vision. Class 8-2 An Application of Clustering. Bernd Heisele 9.913 Class 8-2 An Application of Clustering Bernd Heisele Fall 2003 Overview Problem Background Clustering for Tracking Examples Literature Homework Problem Detect objects on the road: Cars, trucks, motorbikes,

More information

Name: Tutor s

Name: Tutor s Name: Tutor s Email: Bring a couple, just in case! Necessary Equipment: Black Pen Pencil Rubber Pencil Sharpener Scientific Calculator Ruler Protractor (Pair of) Compasses 018 AQA Exam Dates Paper 1 4

More information

Argha Roy* Dept. of CSE Netaji Subhash Engg. College West Bengal, India.

Argha Roy* Dept. of CSE Netaji Subhash Engg. College West Bengal, India. Volume 3, Issue 3, March 2013 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Training Artificial

More information

Chapter 8: Subnetting IP Networks

Chapter 8: Subnetting IP Networks Chapter 8: Subnetting IP Networks Designing, implementing and managing an effective IP addressing plan ensures that networks can operate effectively and efficiently. This is especially true as the number

More information

Joint Entity Resolution

Joint Entity Resolution Joint Entity Resolution Steven Euijong Whang, Hector Garcia-Molina Computer Science Department, Stanford University 353 Serra Mall, Stanford, CA 94305, USA {swhang, hector}@cs.stanford.edu No Institute

More information

Feature Detectors - Canny Edge Detector

Feature Detectors - Canny Edge Detector Feature Detectors - Canny Edge Detector 04/12/2006 07:00 PM Canny Edge Detector Common Names: Canny edge detector Brief Description The Canny operator was designed to be an optimal edge detector (according

More information

Expression Detection in Video. Abstract Expression detection is useful as a non-invasive method of lie detection and

Expression Detection in Video. Abstract Expression detection is useful as a non-invasive method of lie detection and Wes Miller 5/11/2011 Comp Sci 534 Expression Detection in Video Abstract Expression detection is useful as a non-invasive method of lie detection and behavior prediction, as many facial expressions are

More information

Using Layered Color Precision for a Self-Calibrating Vision System

Using Layered Color Precision for a Self-Calibrating Vision System ROBOCUP2004 SYMPOSIUM, Instituto Superior Técnico, Lisboa, Portugal, July 4-5, 2004. Using Layered Color Precision for a Self-Calibrating Vision System Matthias Jüngel Institut für Informatik, LFG Künstliche

More information

GAP CLOSING. Integers. Intermediate / Senior Facilitator s Guide

GAP CLOSING. Integers. Intermediate / Senior Facilitator s Guide GAP CLOSING Integers Intermediate / Senior Facilitator s Guide Topic 3 Integers Diagnostic...5 Administer the diagnostic...5 Using diagnostic results to personalize interventions solutions...5 Using Intervention

More information

Big Ideas of Mathematics, Reception

Big Ideas of Mathematics, Reception Big Ideas of Mathematics, Reception Number Quantities 1, 2 and 3 can be perceptually subitized (recognised as one group without counting) Quantities 4 and 5 can be conceptually subitized (the quantity

More information

Traffic Signal Control Based On Fuzzy Artificial Neural Networks With Particle Swarm Optimization

Traffic Signal Control Based On Fuzzy Artificial Neural Networks With Particle Swarm Optimization Traffic Signal Control Based On Fuzzy Artificial Neural Networks With Particle Swarm Optimization J.Venkatesh 1, B.Chiranjeevulu 2 1 PG Student, Dept. of ECE, Viswanadha Institute of Technology And Management,

More information

Cse634 DATA MINING TEST REVIEW. Professor Anita Wasilewska Computer Science Department Stony Brook University

Cse634 DATA MINING TEST REVIEW. Professor Anita Wasilewska Computer Science Department Stony Brook University Cse634 DATA MINING TEST REVIEW Professor Anita Wasilewska Computer Science Department Stony Brook University Preprocessing stage Preprocessing: includes all the operations that have to be performed before

More information

A MULTI-DIMENSIONAL DATA ORGANIZATION THAT ASSISTS IN THE PARSING AND PRODUCTION OF A SENTENCE

A MULTI-DIMENSIONAL DATA ORGANIZATION THAT ASSISTS IN THE PARSING AND PRODUCTION OF A SENTENCE A MULTI-DIMENSIONAL DATA ORGANIZATION THAT ASSISTS IN THE PARSING AND PRODUCTION OF A SENTENCE W. Faris and K. Cheng Department of Computer Science University of Houston Houston, TX, 77204, USA http://www.cs.uh.edu

More information

Figure 1: Workflow of object-based classification

Figure 1: Workflow of object-based classification Technical Specifications Object Analyst Object Analyst is an add-on package for Geomatica that provides tools for segmentation, classification, and feature extraction. Object Analyst includes an all-in-one

More information

Automated Clustering-Based Workload Characterization

Automated Clustering-Based Workload Characterization Automated Clustering-Based Worload Characterization Odysseas I. Pentaalos Daniel A. MenascŽ Yelena Yesha Code 930.5 Dept. of CS Dept. of EE and CS NASA GSFC Greenbelt MD 2077 George Mason University Fairfax

More information

Optimization of Association Rule Mining through Genetic Algorithm

Optimization of Association Rule Mining through Genetic Algorithm Optimization of Association Rule Mining through Genetic Algorithm RUPALI HALDULAKAR School of Information Technology, Rajiv Gandhi Proudyogiki Vishwavidyalaya Bhopal, Madhya Pradesh India Prof. JITENDRA

More information

Predicting Popular Xbox games based on Search Queries of Users

Predicting Popular Xbox games based on Search Queries of Users 1 Predicting Popular Xbox games based on Search Queries of Users Chinmoy Mandayam and Saahil Shenoy I. INTRODUCTION This project is based on a completed Kaggle competition. Our goal is to predict which

More information

Midterm Examination CS 540-2: Introduction to Artificial Intelligence

Midterm Examination CS 540-2: Introduction to Artificial Intelligence Midterm Examination CS 54-2: Introduction to Artificial Intelligence March 9, 217 LAST NAME: FIRST NAME: Problem Score Max Score 1 15 2 17 3 12 4 6 5 12 6 14 7 15 8 9 Total 1 1 of 1 Question 1. [15] State

More information

CS 231A Computer Vision (Winter 2018) Problem Set 3

CS 231A Computer Vision (Winter 2018) Problem Set 3 CS 231A Computer Vision (Winter 2018) Problem Set 3 Due: Feb 28, 2018 (11:59pm) 1 Space Carving (25 points) Dense 3D reconstruction is a difficult problem, as tackling it from the Structure from Motion

More information

1.2. Pictorial and Tabular Methods in Descriptive Statistics

1.2. Pictorial and Tabular Methods in Descriptive Statistics 1.2. Pictorial and Tabular Methods in Descriptive Statistics Section Objectives. 1. Stem-and-Leaf displays. 2. Dotplots. 3. Histogram. Types of histogram shapes. Common notation. Sample size n : the number

More information

Notes for Unit 1 Part A: Rational vs. Irrational

Notes for Unit 1 Part A: Rational vs. Irrational Notes for Unit 1 Part A: Rational vs. Irrational Natural Number: Whole Number: Integer: Rational Number: Irrational Number: Rational Numbers All are Real Numbers Integers Whole Numbers Irrational Numbers

More information

1. NUMBER SYSTEMS USED IN COMPUTING: THE BINARY NUMBER SYSTEM

1. NUMBER SYSTEMS USED IN COMPUTING: THE BINARY NUMBER SYSTEM 1. NUMBER SYSTEMS USED IN COMPUTING: THE BINARY NUMBER SYSTEM 1.1 Introduction Given that digital logic and memory devices are based on two electrical states (on and off), it is natural to use a number

More information

Heuristic Optimisation

Heuristic Optimisation Heuristic Optimisation Part 10: Genetic Algorithm Basics Sándor Zoltán Németh http://web.mat.bham.ac.uk/s.z.nemeth s.nemeth@bham.ac.uk University of Birmingham S Z Németh (s.nemeth@bham.ac.uk) Heuristic

More information

Introducing Robotics Vision System to a Manufacturing Robotics Course

Introducing Robotics Vision System to a Manufacturing Robotics Course Paper ID #16241 Introducing Robotics Vision System to a Manufacturing Robotics Course Dr. Yuqiu You, Ohio University c American Society for Engineering Education, 2016 Introducing Robotics Vision System

More information

Nearest Neighbor Classifiers

Nearest Neighbor Classifiers Nearest Neighbor Classifiers CSE 6363 Machine Learning Vassilis Athitsos Computer Science and Engineering Department University of Texas at Arlington 1 The Nearest Neighbor Classifier Let X be the space

More information

Pouya Kousha Fall 2018 CSE 5194 Prof. DK Panda

Pouya Kousha Fall 2018 CSE 5194 Prof. DK Panda Pouya Kousha Fall 2018 CSE 5194 Prof. DK Panda 1 Observe novel applicability of DL techniques in Big Data Analytics. Applications of DL techniques for common Big Data Analytics problems. Semantic indexing

More information

LOCALIZATION OF FACIAL REGIONS AND FEATURES IN COLOR IMAGES. Karin Sobottka Ioannis Pitas

LOCALIZATION OF FACIAL REGIONS AND FEATURES IN COLOR IMAGES. Karin Sobottka Ioannis Pitas LOCALIZATION OF FACIAL REGIONS AND FEATURES IN COLOR IMAGES Karin Sobottka Ioannis Pitas Department of Informatics, University of Thessaloniki 540 06, Greece e-mail:fsobottka, pitasg@zeus.csd.auth.gr Index

More information

Supplementary Materials for DVQA: Understanding Data Visualizations via Question Answering

Supplementary Materials for DVQA: Understanding Data Visualizations via Question Answering Supplementary Materials for DVQA: Understanding Data Visualizations via Question Answering Kushal Kafle 1, Brian Price 2 Scott Cohen 2 Christopher Kanan 1 1 Rochester Institute of Technology 2 Adobe Research

More information