/Computer-Vision-Duth

A repo containing all the projects done during the 7th semester university subject "Computer Vision"

Primary LanguageJupyter Notebook

Computer Vision Duth

A repository containing all the projects done during the 7th semester university subject "Computer Vision" The code is written in Python mostly using OpenCV (and for the final project Keras)

Project 1 - Object Counting from a microscope image

In this project we must: Given 2 same images from a microscope (One with added Salt and Pepper Noise)

  • Create a Median filter in order to remove the noise from the second image
  • Count the number of cells that are inside the image boundaries .
  • Count the area (in pixels) of each cell
  • Calculate the average grayscale value for the pixels inside the bounding box of each cell, using a method with constant time complexity (Integral Image)

Project 2 - Image Stitching

In this project we create a Panoramic photo by stitching 4 photos, by extracting their features using the SURF and SIFT algorithms

Project 3 - Image Classification using Bag of Visual Words Model

In this project we classify part of the Caltech-256 Dataset by training a BoV model using k-means and then using the following classifiers:

  • Support Vector Machines (One vs all)
  • K nearest neighbors

For all methods many hyperparameter comparisons were made

Project 4 - Image Classification using Convolutional Neural Netwroks (CNN)

In the final project we classify the same dataset of the previous one using

  • Different Deep Neural Network architectures
  • Data Augmentations
  • Transfer Learning