albertorodrigues / computer-vision-starwars

Computer Vision Project

This is my first computer vision project, necessary to enter the computer vision area of ​​the Artificial Intelligence group at the University of São Paulo called Turing USP. Through it I learned several tools to deal with deep learning and image processing.

The main concepts and tools learned were:

1. Opencv
2. Pytorch
3. Convolutional Neural Network
4. Transfer Learning

Well, no more stalling and let's get down to business! The idea here is to get hands-on with a Computer Vision project to learn the main concepts as the project progresses. Along with this document, I'll leave several links and materials that I used to apply the project's tools.

Main libraries used in the project:

1. Numpy
2. Pandas
3. Matplotlib
4. OpenCV
5. Pillow
6. PyTorch
7. Sklearn
8. Scikit-image
9. glob
10. os

PyTorch was the main library used for deep learning while the others were mainly in image manipulation, image files and image visualization.

There are two parts to this project:

1. Image manipulations

• In this step, I showed my mastery over the images: reading, manipulating the RGB bands and even applying transformations over the images.

• For example, considering the following original image:

  

• Some manipulations and transformations performed were:

  

• Are you curious? Click here to go straight to this part.

2. Deep Learning

• Now comes the biggest challenge in this project, which is deep learning.

• We apply some algorithms to classify three classes of images:

  1. Yoda

     

  2. Stormtrooper

     

  3. Darth Vader

     

  We will use two methods for this purpose:

  1. Principal Component Analysis and then Logistic Regression

  2. Convolutional Neural Network

  

• Click here to go straight to this part.