In this chapter, we saw several algorithms that perform face alignment. Right at the outset, we stated the importance of such alignment steps in a processing pipeline. Especially when we are working with faces, aligning them leads to significant enhancements in overall accuracies.
We started off by running the face detection code from the previous chapter on the face images in our dataset. After that, we used affine transformations to scale, crop, and rotate our face images. By the end of the sequence of operations, we had transformed our set of faces into a form that we claim to be very useful for our subsequent step--feature extraction.
Throughout the chapter, we also showed some qualitative results of running the various algorithms on images taken from the dataset. By now, you should have developed an intuitive sense of what each affine transformation accomplishes and how these changes manifest in the output images.
In the next chapter, we will be working with the transformed and...