reu report i

15
REU REPORT I Alla Petrakova UCF

Upload: alka

Post on 20-Jan-2016

34 views

Category:

Documents


0 download

DESCRIPTION

REU Report I. Alla Petrakova UCF. Material covered. MATLAB Derivatives, Filters, Thresholding , Equalization, etc. Correlation, Convolution Edge Detection ( Sobel , Laplacian of Gaussian, Canny) Harris Corner Detector SIFT Adaboost , face detection SVM Optical Flow Bag of Features. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: REU  Report I

REU REPORT IAlla PetrakovaUCF

Page 2: REU  Report I

MATERIAL COVERED

MATLAB Derivatives, Filters, Thresholding, Equalization, etc. Correlation, Convolution Edge Detection (Sobel, Laplacian of Gaussian, Canny) Harris Corner Detector SIFT Adaboost, face detection SVM Optical Flow Bag of Features

Page 3: REU  Report I

EDGE DETECTION - SOBEL

Page 4: REU  Report I

GAUSSIAN

Page 5: REU  Report I

EDGE DETECTION - SOBEL

Page 6: REU  Report I

LAPLACIAN OF GAUSSIAN

Page 7: REU  Report I

CANNY

Page 8: REU  Report I

HARRIS CORNER DETECTOR

Page 9: REU  Report I

HARRIS CORNER DETECTOR

Page 10: REU  Report I

SIFT – DENSE SAMPLING

Page 11: REU  Report I

SIFT – DESCRIPTORS INCLUDED

Page 12: REU  Report I

OPTICAL FLOWCOMPARING OUTPUTS – CE LIU

Page 13: REU  Report I

OPTICAL FLOW

Window size = 70

Window size = 40

Page 14: REU  Report I

SVM & BAG OF WORDS

SVM One of the biggest challenges Tried with sift, dense sift, scaled data Stubbornly stuck on 53% accuracy

Bag Of Words 47% to 53% accuracy

Possible solution: “A Practical Guide to Support Vector Classification Chih-Wei Hsu, Chih-

Chung Chang, and Chih-Jen Lin “ http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf

Page 15: REU  Report I

RESEARCH PROJECTS

“Trajectory Clustering: A Motion Pattern approach” by Mahdi M. Kalayeh Pattern recognition Analysing effects of applying various similarity measures Probabilistic Predictive Modelling

“Clustering in High Dimensional Data” by Gonzalo Vaca-Castano

“Cell Tracking and Lineage Construction” with Sarfaraz Hussein