This repo consists of some basics of OpenCV & Computer Vision. Also, there are 3 projects in this repo, namely
- Virtual Paint 🖌️
- Document Scanner 🖨️
- Number Plate Detector 🚗
- Chapter 1 - Importing an image, a video, or to get live feed using a Webcam with the help of OpenCV Library
- Chapter 2 - Applying various image transformations such as Chaning Color Schemes (RGB to Gray, etc), Blur (Gaussian Blur), Egde Detection, Thickening and Thinning of Edfes
- Chapter 3 - Resizing and Cropping Images
- Chapter 4 - Drawing Shapes such as Circles, Rectangles, Lines, etc on Images, and putting text on Images
- Chapter 5 - Selecting a particular part of an image, & displaying it, after transforming it's Perspective.
- Chapter 6 - Stacking several images together in a single Window
- Chapter 7 - Obtaining masks out of an Image, by selecting ranges for Hue, Saturation & Brightness
- Chapter 8 - Classification of Shapes on the basis of Edge Detection & Contours
- Chapter 9 - Face Detection using Haar Cascades
- In this project, with the help of what we learnt in Chapter 7, we are detecting the masks of 3 different colors, namely Green, Orange & Purple.
- We are varying the min & max values of Hue, Saturation & Brightness, to perform the above task.
- If we are detecting any color in a given frame, then we are finding it's Contours, and calculating the area of that Contour.
- If the area is bigger than the Minimum Area (500) in this case, then we are storing the Middle, Top-point Coordinates of the Contour in an array, along with an Index to represent the color.
- And finally, we are iterating over the array, and plotting a point with the respective color, on the frame.
- A Live Demo of this project, can be checked out on Youtube
- In this project, we are trying to detect documents in the frame, and are trying to select it from the entire frame, similar to what we did in Chapter 5.
- The first step is the Pre-Processing of the image, which includes conversion to Gray Color, using a Gaussian Blur, Edge-detection, and then Dilation & Erosion of the edges.
- We have done Dilation & Erosion, to increase the stability & reliability of detection.
- After that, we are trying to get the Contours of the different objects present in the frame, and finding out the object with the Biggest Contour in the frame, which we are assuming as our Document.
- And then we are just transforming the Perspective of that object, to select it from the entire frame, and display it separately.
- The code can work on both Live Feed as well as Static Images, with slight modifications (Mentioned in the code itself). You can see an example of it below 👇
- In this project, we are detecting a Number Plate out of a Static Image or Live Feed, using Haar Cascades, similar to what we learnt in Chapter 9.
- If we want to, we can also save the detected Number Plate, in an Image Format.
- You can see it working & some examples of it below 👇
