MATLAB: Computer Vision and GPU Acceleration for new laptop selection

Question

Hello, we are buying a ~$2k laptop for MATLAB stereo vision processing of dual 1920×1080 streams.

We will be doing real-time object detection with these cameras, ground truth database creation, and other processing tasks. Computer Vision is our main toolbox that we use and we want to know if getting a super high end GPU option is really necessary.

We are currently looking at laptop units with the i7 6700HQ @ 2.6GHz to 3.5GHz Max (4cores, 8threads), 16GB DDR4 RAM (enough?), and 512GB PCIe SSD as the base specs. We are however stuck on if we should get a laptop with the Nvidia 960M or the new mobile GTX 1060?

I know that MATLAB supports CUDA and uses acceleration for the image processing toolbox and various others.

I just don't know if the performance boost or acceleration it provides is enough to warrant getting a laptop with the new mobile GTX 1060 with 6GB VRAM? And if the VRAM of the GTX 960M 2GB vs 4GB matter to MATLAB. Even a 20% boost in general performance is enough for us to go with the GTX 1060 6GB. But we want to check with you all first.

I appreciate any help you may have and if you have suggestions.

Thank you.

Answer 1

The Computer Vision System Toolbox itself does not support too many functions on the GPU. The workflows that require GPU computing as of R2016b in computer vision are deep learning, like in this example:

Image Category Classification using Deep Learning

However, a lot of functions in MATLAB and Image Processing Toolbox itself support GPU computing, which you may benefit from:

GPU enabled Image Processing Functions

GPU enabled MATLAB Functions