Getting StartedSession 1

Computer Vision GroupIIT Madras

December 1, 2014

(Computer Vision Group) Getting Started December 1, 2014 1 / 18

Outline

1 Installation OpenCVWindowsUbuntu

2 Concepts in Image ProcessingPixelsImage ProcessingImage Transformation

3 Feature ExtractionWhat are features?Some feature extraction tools

(Computer Vision Group) Getting Started December 1, 2014 2 / 18

Installing OpenCVWindows 7/8

Install Anaconda 2.0.1

Download OpenCV 2.4.9 and extract to a convenient location

Go to opencv/build/python/2.7 folder

Copy cv2.pyd to INSTALL DIRECTORY/Python/lib/site-packages

Open IPython QT console.

import cv2

If you don’t get any errors, its a success.

(Computer Vision Group) Getting Started December 1, 2014 3 / 18

Installing OpenCVUbuntu

Using the apt-get tool# sudo apt-get install python-opencv

Installs an older version of OpenCV. Build from source to get the latestversion.

(Computer Vision Group) Getting Started December 1, 2014 4 / 18

PixelsRecap

Basic building blocks of an image

Color represented as a tuple (R, G, B)

(Computer Vision Group) Getting Started December 1, 2014 5 / 18

Image ProcessingRecap

Thresholding

Erosion

Dilation

(Computer Vision Group) Getting Started December 1, 2014 6 / 18

ThresholdingImage Processing

(Computer Vision Group) Getting Started December 1, 2014 7 / 18

ErosionImage Processing

(Computer Vision Group) Getting Started December 1, 2014 8 / 18

DilationImage Processing

(Computer Vision Group) Getting Started December 1, 2014 9 / 18

Image TransformationFairly Simple

Rotation

Translation

Cropping

Warping

(Computer Vision Group) Getting Started December 1, 2014 10 / 18

RotationImage Transformation

(Computer Vision Group) Getting Started December 1, 2014 11 / 18

TranslationImage Transformation

(Computer Vision Group) Getting Started December 1, 2014 12 / 18

WarpingImage Transformation

(Computer Vision Group) Getting Started December 1, 2014 13 / 18

Feature ExtractionWhat are features?

Feature Extraction in Images

Transforming rich content of images into a set of values. Featureextraction is a crucial part in Machine Learning.

Example

Histograms are commonly used for extracting set of features. More featureextraction techniques coming up.

Digit Recognizer

We’ll be using Machine Learning to build a digit recognizer in tomorrow’ssession.

(Computer Vision Group) Getting Started December 1, 2014 14 / 18

Feature ExtractionWhat are features?

Feature Extraction in Images

Transforming rich content of images into a set of values. Featureextraction is a crucial part in Machine Learning.

Example

Histograms are commonly used for extracting set of features. More featureextraction techniques coming up.

Digit Recognizer

We’ll be using Machine Learning to build a digit recognizer in tomorrow’ssession.

(Computer Vision Group) Getting Started December 1, 2014 14 / 18

Feature ExtractionWhat are features?

Feature Extraction in Images

Transforming rich content of images into a set of values. Featureextraction is a crucial part in Machine Learning.

Example

Histograms are commonly used for extracting set of features. More featureextraction techniques coming up.

Digit Recognizer

We’ll be using Machine Learning to build a digit recognizer in tomorrow’ssession.

(Computer Vision Group) Getting Started December 1, 2014 14 / 18

Feature Extraction ToolsThere are many more available

1 Binarized pixel values

2 Intensity histogram

3 Histogram of Oriented gradients

4 SIFT

(Computer Vision Group) Getting Started December 1, 2014 15 / 18

Intensity HistogramFeature Extraction

(Computer Vision Group) Getting Started December 1, 2014 16 / 18

Histogram of Oriented GradientsFeature Extraction

(Computer Vision Group) Getting Started December 1, 2014 17 / 18

Summary

Today’s session

Image processing/transformations

Feature Extraction

Tomorrow’s session

Machine Learning Basics

Training a classifier for handwritten digit recognition

(Computer Vision Group) Getting Started December 1, 2014 18 / 18