Overview

This program offers various functionalities to process images & videos using OpenCV. These functionalities include transformation of image & video to grayscale, edge detection, Gaussian blurring, thresholding, motion tracking, and object detection.

Requirements

1. C++20 Compiler
2. CMake: https://cmake.org/install/
3. OpenCV: Version 4.6.0_1. Download and Install the latest version of OpenCV for your respective machine from https://opencv.org/releases/

Setup

For macOS:

Firstly, install homebrew from https://brew.sh/.

1. Install OpenCV: Run brew install opencv. Check that brew has been installed at /opt/homebrew/Cellar/opencv/4.6.0_1/lib/pkgconfig/opencv4.pc

Alternatively, do the following:

1. Extract OpenCV from the downloaded file.
2. Move OpenCV to the proper library directory: /opt/homebrew/Cellar/opencv/4.6.0_1/lib/pkgconfig/opencv4.pc

For Both Windows and macOS:

1. Extract our project from the Github repository or a zip file. https://github.com/vicentefarias/4995_project/tree/Final
2. Change into the project directory.
3. Run make from the project directory and ensure the code compiles (this may require checking library path directory in Makefile).

Usage

Create a Image or Video object using a file path or an instance of cv::VideoCapture. After initializing the Video object, you can call any of the following methods to apply a specific transformation to the them:

• grayscale: converts the image/video to grayscale
• edge_detect: applies edge detection to the image/video
• gaussian_blur: applies Gaussian blurring to the image/video
• threshold: applies thresholding to the image/video
• track: tracks an object in the image/video using OpenCV's KCF tracker
• detection: detects objects in the image/video using YOLOv5 object detection model

Documentation

Please find the documentation.pdf included.

Benchmarking

We were interested in determining whether our wrapper for OpenCV would be faster than the Python code using OpenCV, which actually runs C++ in the background. To measure the time taken by each function call, we created a Benchmark.cpp file. We also created a benchmark script Benchmark.py written in Python to compare the efficiency of our implemented library against the Python code using OpenCV. We created an Image class that imitates the Image class we created in C++.

Here are some comparison of some functionalities on our project against the OpenCV functions through Python):

	Edge Detection	Gaussian Blurring	Alpha Blend
Our library & utility	2.58	2.88	1.48
OpenCV with Python	2.67	3.04	1.89
Results we are	3% faster	5% faster	22% faster

Special Thanks to Dr. Bjarne Stroustrup

As students working on this project, we were incredibly fortunate to have the opportunity to be mentored by such an esteemed figure.

Throughout the project, Dr. Stroustrup provided invaluable guidance on a range of topics, including software design, code optimization, and debugging techniques. His extensive knowledge and expertise allowed us to learn practical skills that are not typically covered in a traditional classroom setting. We often found ourselves surprised at the depth and breadth of his insights, and we were grateful for the opportunity to learn from one of the industry's most seasoned experts.