Taking the landmark points customization to the next level, we will use the landmark points to create a custom face make-up for the face image.
Till now we were using the face-recognition third party library to achieve most of the functionality. But from now onwards we will try the face-recognition pipeline steps which includes face detection, face alignment, face feature extraction verification and classification separately one by one using popular libraries. We will have an introduction about these in this session.
In the next session, we will start with face detection. We will divide them into traditional face detection methods and modern methods which involves CNN.
At first we will try the Haar Cascade object detection algorithm for face detection. We will try it at first for still images and later we will implement it for saved videos as well as live web cam videos.
Another popular algorithm for face detection is HOG or Histogram of Oriented Gradients. At first we will have an introduction to the working of HOG algorithm and then we will try the HOG method for images, videos and real-time web cam stream.
The next face detection algorithm we will try is SSD or Single Shot Detection. We will repeat the same functionality exercises for SSD also. And then comes MMOD. We will repeat the same functionality exercises for SSD also.
Then the MMOD, the Max-Margin Object Detection. We will repeat the same functionality exercises for MTCNN also.
Then comes the next algorithm which is MTCNN, the Multi-task Cascaded Convolutional Networks. We will repeat the same functionality exercises including image, video and real time stream for MTCNN also.
Finally we will have a quick comparison between the performance of these face detection algorithms.
After face detection, we will go ahead with face alignment. We will use the popular Dlib library python implementation to perform the face alignment for image, video and video streams.
After face alignment exercises, we will proceed with face verification and classification where the actual face recognition is happening. At first we will have an introduction about face classification. We will divide the techniques into traditional face recognition methods and modern methods which involves CNN.
At first we will try the techniques Eigenface Fisherface and LBPH, the Local Binary Pattern Histogram. We will have a short introduction about these algorithms and will then proceed with
preparing the image dataset for these algorithms.
Then we will set up the prerequisite for them. Later we will proceed with face detection using MTCNN and preprocessing of the detected face for recognition. Then the exercises involving training with the image dataset and trying prediction for images. We will then save this model so that we can load it later and do prediction without having to go through training again.
We will also try it for pre-saved videos and real time webcam stream. Once we are done with that we will have a quick comparison of the Eigenface Fisherface and LBPH algorithms.
That’s all the traditional ways, now we will proceed with deep learning face recognition. At first using the popular VGGNet model for face recognition called VGGface. We will have an introduction to VGG face and then we will implement VGGface face verification for images. Later we will try VGGface face verification for videos as well as realtime streams.
And then we have an introduction to FaceNet, OpenFace and DeepFace Models. We will use a popular easy to use open source python face recognition framework called deepface to implement the rest of popular deep learning techniques.
We will install deepface to our computer and then try it at first for face detection and face alignment. Then we will try deepface for face one to one verification. Later with few changes, we can use it for face classification which involves an one to many comparison. deepface can also be used for performing face analysis involving gender, age, emotion, ethnicity etc
That’s all about the topics which are currently included in this quick course. The code, images and libraries used in this course has been uploaded and shared in a folder. I will include the link to download them in the last session or the resource section of this course. You are free to use the code in your projects with no questions asked.
Also after completing this course, you will be provided with a course completion certificate which will add value to your portfolio.
So that’s all for now, see you soon in the class room. Happy learning and have a great time.
Who this course is for: