Pattern Recognition
Total Credits : 03 Subject Code : PECS6T003
Teaching Scheme Examination Scheme
Teaching Hrs/Week : 03 Duration of End Semester Exam : 03 Hrs.
Tutorial/Activity Hrs/Week : 00 Internal Evaluation : 40 Marks
Practical Hrs/Week : 02 End Semester Examination : 60 Marks
Course Objectives
To provide a foundational understanding of the core principles, mathematical models,
1.
and design cycle of pattern recognition systems.
To equip students with practical skills to pre-process data, apply fundamental
2. classification and clustering algorithms, and evaluate model performance.
Course Outcomes
After completion of course, students would be able to
CO1 Explain the fundamental concepts of pattern recognition, including the design cycle, learning
approaches, and the principles of Bayesian decision theory.
CO2 Apply various pre-processing and feature extraction techniques to prepare datasets for analysis.
CO3 Implement and compare basic supervised learning algorithms
CO4 Utilize unsupervised learning algorithms like k-Means, Hierarchical, and Fuzzy c-Means
clustering to discover hidden structures in data.
CO5 Describe the basic architecture and function of simple Neural Networks (MLP, CNN) for
recognition tasks.
CO6 Evaluate the performance of pattern recognition models using metrics.
BOS Chairperson Dean (Academic) Principal
Pattern Recognition
COURSE CONTENT
UNIT-I
Introduction to Pattern Recognition: Definition, need, and applications of pattern recognition, Design cycle
and components of a pattern recognition system, Supervised, Unsupervised and reinforcement learning
approaches, Statistical, Structural and neural network-based pattern recognition methods, Bayesian
decision theory and probability.
UNIT-II
Feature Extraction and Data Pre-processing: Data collection, Normalization and transformation, Feature
selection vs feature extraction, Dimensionality reduction- PCA, LDA, Distance and similarity measures
(Euclidean), Handling noise, Missing data and outliers.
UNIT-III
Supervised learning: k-NN, Pros and cons, Bayesian principle for classification, Gaussian Naïve Bayes
concept of information entropy, Decision tree, Intuitive concept of a maximum-margin hyperplane, linear
SVM.
UNIT-IV
Unsupervised learning: Clustering, Types of clustering, k-Means, Hierarchical, Agglomerative approach,
understanding dendrograms, Fuzzy c-Means,
UNIT-V
Neural networks and performance evaluation: Perceptron, Multi-Layer Perceptron (MLP), Neural network
classifiers, CNN basics for recognition, Performance metrics- confusion matrix, Precision, Recall, F1
score, Evaluation, Validation and model improvement techniques.
BOS Chairperson Dean (Academic) Principal
Text books:
1. Christopher M. Bishop, Pattern Recognition and Machine Learning, Springer Pub.
2. Ethem Alpaydin, Introduction to Machine Learning, MIT Pub.
Reference books:
1. R.O. Duda, P.E. Hart, D.G. Stork, Pattern Classification, Wiley Pub.
2. Sergios Theodoridis, Konstantinos Koutroumbas , Pattern Recognition, Academic Pub.
Virtual Reality and Augmented Reality
Total Credits : 01 Subject Code : PECS6P003
Teaching Scheme Examination Scheme
Teaching Hrs/Week : 00 Duration of End Semester Exam : --
Tutorial/Activity Hrs/Week : 00 Internal Evaluation : 25 Marks
Practical Hrs/Week : 02 End Semester Examination : 25 Marks
Course Objectives
To provide a foundational understanding of the core principles, mathematical models,
1.
and design cycle of pattern recognition systems.
To equip students with practical skills to pre-process data, apply fundamental
2. classification and clustering algorithms, and evaluate model performance.
Course Outcomes
After completion of course, students would be able to
CO1 Explain the fundamental concepts of pattern recognition, including the design cycle, learning
approaches, and the principles of Bayesian decision theory.
CO2 Apply various pre-processing and feature extraction techniques to prepare datasets for analysis.
CO3 Implement and compare basic supervised learning algorithms
CO4 Utilize unsupervised learning algorithms like k-Means, Hierarchical, and Fuzzy c-Means
clustering to discover hidden structures in data.
CO5 Describe the basic architecture and function of simple Neural Networks (MLP, CNN) for
recognition tasks.
BOS Chairperson Dean (Academic) Principal
Pattern Recognition
Perform any 10 Experiments.
1. Install and explore Python environment for pattern recognition (OpenCV, scikit-learn, numpy).
2. Load a dataset (e.g., Iris or digits) and visualize patterns using scatter plots.:
Implement z-score normalization and min–max scaling.
3. Apply Principal Component Analysis on a dataset and reduce dimensions.
4. Train and test a k-Nearest Neighbour classifier on the Iris dataset.
5. Classify data using a linear SVM and visualize decision boundaries.
6. Cluster an image or dataset using k-means algorithm.
7. Implement a simple MLP (Multi-Layer Perceptron) using sklearn for digit recognition.
8. Create a basic 3D virtual environment (e.g., simple room or object).
9. Enable simple camera movement or object interaction (keyboard/mouse control) in Unity.
10. Implement real-time face or hand detection using Cascades.
11. Implement Object Detection using a Pre trained Deep Learning Model (e.g., MobileNet/ YOLO /
SSD).
12. Integrate Augmented Reality (AR) Elements in a Unity Scene.
* Above is the recommended list but not limited to this.
BOS Chairperson Dean (Academic) Principal