Computer Vision Project

Projects of the subject Computer Vision at the University of Granada (UGR). Course 2021/2022

Contents

Assignment 0

Image reading with OpenCV, Numpy and MatPlotLib.

Assignment 1

Manual implementation of different tools used in Computer Vision:

• Approximation of Gaussian mask and its 1st and 2nd derivative masks.
• Convolution using separable masks.
• Construction of Gaussian and Laplacian pyramids.

In addition, two applications of the above are added:

• Reconstruction of an image from its Laplacian pyramid.
• Creation of hybrid images (greyscale and colour).

Assignment 2

• Implementation of the SIFT algorithm for characteristic points detection
• Construction of panoramas using Homographs

Assignment 3

Use of convolutional neural networks.

Final Proyect

Use of the ResNet50 model to perform Transfer Learning (using such a network previously trained on ImageNet) in order to assist in the diagnosis of pulmonary pneumonias in X-ray images. Other different models are also tested to discern which one provides the best results for this problem.

Tech Stack

Languages: Python 3.7

Libraries: OpenCV, Numpy, Keras, Matplotlib

Others: Google Colab