OpenCV Python Development: Detecting Faces and Features

Name: OpenCV Python Development: Detecting Faces and Features
Author: Tom Rosenwinkel

Learn how to identify features and faces in OpenCV

Preview the first lesson free — get full access to all 9 lessons.

Course: On-Demand

Hello, in this lesson, we're going to see examples of template matching in OpenCV. As we begin detecting features in images, we should first ask what makes a good feature, and in template matching, a feature is template image itself. You'll see that I've specified the path to an example image in a raw subdirectory in my project directory. They're finding the file at that path and opening it and writing the raw version to a subdirectory called out. But I've also drawn a rectangle to indicate where a basketball is within this image and written that as well as the sub image of that basketball itself. We will use these to do template matching. So to see the example image, the raw image has a basketball in it. We've found the basketball and the basketball itself looks like this. Now if we did not already know where that basketball was in the image, we can find it using OpenCV's match template function. It will create a score for each pixel that is an indication of the likelihood that that basketball or template image might be at that pixel. The arguments for the match template function are the image we are searching and the template image in this case, the basketball, and then a flag to indicate what method to use. OpenCV has released a list of these methods. They all begin with TM underscore. For this first example, I'll use see the cross coefficient normed. This cross coefficient compared with the other one released by OpenCV is a little better shaped. We can look at what these match scores look like because they are normed. They will be between 0 and 1, so we can see them if we scale to 255. So we can see on the scores as an image, the dark places are places where the basketball is unlikely to be found, and the bright places are where it is more likely to be found. The brightest place is actually where the basketball was in the original image roughly, and we can mark that on our image by drawing a rectangle for that place. To locate the place, we can use OpenCD's min max loc function, which will return the values and locations of both the max and the min. To draw the rectangle using those coordinates, we will remember that the rectangle annotation function requires that we specify the lower corner of the image and that to calculate the upper corner, we will need to add the size of the basketball to that location. We will draw this box in green, so that's our green rectangle that we drew. Another method that is commonly used in OpenCV would be the square root diff method. We will again use the normed version, and as a square root of the differences, we should remember that the lower scores indicate a better match for the template, and so when we draw our box, we should not use the maximum score, but the minimum score. So again, here for the square root diff, the scores, the darker ones indicate the places where the basketball is most likely to be, and so there's our green box. In the next lesson, we will look at the case where the template image might appear multiple times within an image, and so we will not just want to find the max or min score, but we will want to threshold those scores. Thank you for watching.

Intermediate Provider : Tom Rosenwinkel 9 Lessons · 44m in English

Course Description

How can you locate particular objects in images by their features? How can you apply advanced classical and machine learning technology to solve your computer vision problems? OpenCV is a well-established open source project that makes computer vision available to any Python developer. Learning how to process images and videos and extract features relevant to your project is an essential skill in many data and application projects. The topics discussed in these lessons will provide a foundation for future work in computational image enhancement or computer vision AI applications.

In these lessons on face and feature detection, you will learn how to detect particular objects in images and video frames in OpenCV. We will explore template matching, corner detection, more advanced feature detection and matching, and classical and machine learning approaches. By the end of these lessons, you will be able to detect faces and a variety of objects by their features in OpenCV.

What You'll Learn

Detect particular objects in images and video frames using OpenCV
Apply template matching, including template matching with thresholding
Perform corner detection and advanced feature detection with SIFT, SURF, FAST, and ORB
Match features between images to identify and recognize objects
Use Haar cascading classifiers and train classifiers for detection
Detect faces and eyes by applying classical and machine learning approaches

Key Takeaways

OpenCV is an open source project that makes computer vision available to any Python developer.
The course covers both classical and machine learning approaches to computer vision problems.
Learners explore template matching, corner detection, and advanced feature detection and matching.
By the end of the lessons, learners can detect faces and a variety of objects by their features in OpenCV.
These lessons provide a foundation for future work in computational image enhancement or computer vision AI applications.

Frequently Asked Questions

What will I be able to do after completing this course?

By the end of these lessons, you will be able to detect faces and a variety of objects by their features in OpenCV.

What topics does this course cover?

It covers template matching, corner detection, more advanced feature detection and matching, and classical and machine learning approaches, including Haar cascading classifiers and face and eye detection.

What skills will I gain from this course?

You will build skills in computer vision, face detection, feature detection, object detection, object recognition, and OpenCV.

Who is this course for?

It is for Python developers who want to process images and videos and extract features relevant to data and application projects using OpenCV.

What feature detection methods are taught?

The lessons cover SIFT, SURF, FAST, and ORB, along with feature matching and corner detection in OpenCV.