Machine Learning for
Computer Vision
Unit-VI
Computer Vision
Introduction to Machine Learning
Image Analysis (Computer Vision)
Fundamental Steps
• Preprocessing (Image enhancement and Image reconstruction)
• Segmentation (Isolating objects)
• Feature Extraction (Representation and description)
• Classification (Object recognition)
Introduction to Machine Learning Contd.
Image Classification & Recognition
What is pattern recognition?
"The assignment of a physical object or event to one of several prespecified categories" —
Duda & Hart.
• A pattern is an object, process, or event that can be given a name.
• A pattern class (or category) is a set of patterns sharing common attributes and usually
originating from the same source.
• During recognition (or classification), given objects are assigned to prescribed classes.
• A classifier is a machine which performs classification.
Pattern Recognition (PR)
• Sensors and preprocessing
• Feature extraction aims to create discriminative features good for classification
• Classifier
• Teacher provides information about hidden state — supervised learning
• Learning algorithm sets PR from training example.
Example
Feature Extraction
Extract features that are good for classification.
Good features:
• Objects from the same class have similar feature values.
• Objects from different classes have different values.
Machine Learning Methods
• Recurrent Neural Net
• Convolutional Neural Net
• Neural Net
• Deep (sparse/denoising) Autoencoder
• Deep Belief Net
• BayesNP
• Boosting
• Perceptron
• SVM
• Autoencoder Neural Net
• Sparse Coding
• GMM
• Restricted BM
Machine Learning Methods Contd.
Machine Learning Methods Contd.
Supervised: Learning with a labeled training set
Example: email classification with already labeled
emails
Unsupervised: Discover patterns in unlabeled data
Example: cluster similar documents based on text
Reinforcement Learning: Learn to act based on
feedback/reward
Example: learn to play Go, reward: win or lose
Machine Learning Basics
Comparision
Classifier example -- SVM (Support Vector Machine)
• For Linearly Separable dataset
• Hyperplane creation yi w xi b 1
• All negetive yi in one side of hyperplane and all positive on other side.
• For Non-separable dataset
• Use of slack variable i 0, represents the amount by which constraint is violated.
• So, yi w xi b 1 i
Dimension reduction--PCA (Principal Component Analysis)
• For high-dimensional feature
vectors, it is more difficult to
estimate a classifier accurately,
because it will tend to increase
the variance of the estimate.
• It can be useful to reduce the
dimension of a feature vector. We
can do so by projecting onto a
low-dimensional basis.
• D ro p t h e fe at u re s t h at h ave
relatively little variance.
Object Detection
• Sliding Window Method
• Face detection
• Human detection
• Boundary detection
• Detecting deformable objects
Computer vision and machine learning for applied
research
• Optical Character Recognition (OCR)
• Handwritten: sorting letters by postal code, input device for PDAs
• Printed texts: reading machines for blind people, digitalization of text documents
• Biometrics
• Face recognition, verification, retrieval
• Fingerprints recognition
• Speech recognition
• Diagnostic Systems
• Medical diagnosis: X-Ray, ECG analysis
• Machine diagnostics
• Military Applications
• Automated Target Recognition (ATR)
• Image segmentation and analysis (recognition from aerial or satellite photographs)