CNN have pushed the boundary in image classification, segmentation and detection but that is not say that is all there is to computer vision.

Additionally, there are some problems which have better solutions using

traditional techniques with global features (I know one of them in my practices).

On the other hand, knowledge is never obsolete. This changes open many doors to do something in those traditional techniques to improve in differents aspects (e.g. compute power, time, accuracy, characteristics and quantity of inputs, and among others).

Finally, There are many more changing problems in computer vision such as: Robotic, augmented reality, automatic panorama stitching, virtual reality, 3D modelling, motion stamation, video stabilisation, motion capture, video processing and scene undertanding.

In one word I can say No. Deep learning is used in the domain of Digital Image Processing in order to solve some problems (Ex. to color gray-scale videos). You can check the following link:

https://www.analyticsvidhya.com/blog/2017/02/6-deep-learning-applications-beginner-python/?utm_content=bufferba7cb&utm_medium=social&utm_source=linkedin.com&utm_campaign=buffer

But in any case we have a lot of problems in DIP to be solved using the other "traditional" techniques.

I think No. Standard technique is still used when you have low processing power and when you are not in a big data like problem.

I believe no just like other researchers.

CNN have pushed the boundary in image classification, segmentation and detection but that is not say that is all there is to computer vision.

Additionally, there are some problems which have better solutions using

traditional techniques with global features (I know one of them in my practices).

On the other hand, knowledge is never obsolete. This changes open many doors to do something in those traditional techniques to improve in differents aspects (e.g. compute power, time, accuracy, characteristics and quantity of inputs, and among others).

Finally, There are many more changing problems in computer vision such as: Robotic, augmented reality, automatic panorama stitching, virtual reality, 3D modelling, motion stamation, video stabilisation, motion capture, video processing and scene undertanding.

Off course not

Deep Learning methods such as CNN mostly improve prediction performance with big data sets. According to my knowledge, millions of data records are often required. When big datasets or high computing facility are unavailable , traditional methods will come into play.

DL is doing very good for solving a number of image/video related problems. But, one can boost the power of DL using image processing/traditional techniques. One has to put image processing techniques in several layers of DL when he/she is not using CNN/LSTM etc. as black box.