Regardless the impact machine learning techniques are having on the field, I still think is valuable to know and understand what are the classic methods that they are allegedly replacing.

It is actually not about "classic" vs. "modern". A lot of what's described in the book is as relevant as ever.

Mostly what's changed is that what used to be heuristics and hand-crafted features is now getting replaced by proper learned models. Also some generative models are getting replaced discriminative models since deep learning does a very good job in creating those.

Sure. Just it's nice to know which pieces are:

- considered obsolete (e.g. complicated and nasty, and still outclassed by NNs),

- building blocks of current techniques,

- provided inspiration for current techniques,

- are still state of the art.

A lot in the book is about geometry and Structure from Motion. It's an area that's becoming augmented but not superseded by Neural Nets. For example if you want to generate a 3D model from images you can't just throw a NN at it. Same for tracking etc...

Although that might change real soon. There are a number of ongoing works that attempt at using CNNs to estimate both geometry and movement.

Got a link?

SLAM using SfM (from monocular vision) using Deep Nets would be huge. Would reduce the specifications for LIDAR in SDC's for example.

Things like principal component analysis, eigenvalues and filtering are timeless and fundamental tools that won't go away in the near future.

Even if you don't go into computer vision - these have a wide range of applications, for example in sound or signal processing which are closely related.

I used this book in my digital image processing class (a class for EE majors). A lot of the book is applications of classic signal processing techniques.