A University AI Course, practicals, code and examples
##Practical Part 1 ###1 Goal Given a binary image, write a code in C++ that uses Depth First Search algorithm to find all the connected components in the image. For each connected component find:
- the centroid
- the 7 invariant moments
- the perimeter and the surface area
Create a file in which each record containsdata collected from each image(number of connected components, data collected in (a), (b) and (c)) ###2 The dataset The data set used consists of 20 folders each containing 3 images(60 images in total).
##Practical Part 2 ###Goal Artificial Neural Networks(ANN) is a popular and effective machine learning solution to many classification problems. In this project, You will learn the different steps that are needed to design and use ANN. It is then used to classify animals using their footprints. Data sets are provided. ###2 The data set The data set used consists of images of footprints of animals (Baboon, Buffalo, Cheetar, Elephant, etc.). ###3 Design and Implement an Artificial Neural Network to classify Animals using their footprints. *1. Matlab uses separate files for the input patterns and the output patterns. The inputs must be stored to one file and outputs to another. Make sure that you don’t mess up the order of records across the two files. *2. Study the MATLAB Neural Networks toolbox, and split the file into training, validation and test sets. Read the documentation and see how this should be done in Matlab and how the files should be named. *3. Design and implement a Multilayer perceptron and use it to classify animals represented by the files given above. Two experiments must be carried out: � Experiment 1: inputs are the pixels of the image of an animal footprint. � Experiment 2: inputs are the features extracted from an image of animal footprint generated in practical 1. ###4 Performance evaluation Compute how many of the animals are correctly classified and also how many are wrongly classified for each class, in each of the experiments carried out in section 3. Build the confusion matrix, and calculate True Positive and the False Positive rates [1]. Compare the classification performances of the two experiments.