• Good article but not 100% depth researched.. Itseez, for example, started off as an offshoot of Intel.. OpenCV started at Intel's Nizhny Novgorod lab. The team in Russia developed into Itseez.. It's founders (Alexander Bovyrin, Victor Erukhimov, Sergey Molinov) all worked for Intel on the library.. Willow Garage (Menlo Park), especially Brian Gerkey, picked it up.. and the quality grew. OpenCV is quite popular and hands-down the most popular computer vision platform-- it's my chosen platform. More interesting was the acquisition of RealSense (an Israeli start-up whose founder is currently doing work on 3D Deep Learning). Where NVIDIA has their stength is the use of CUDA for "Deep Learning" training. NVIDIA created CUDA and is clearly responsible for the rebirth of ANNs-- putting ANNs on GPUs was the trick. They own the territory. Distruption shall come in the short term from the domain of low power inference-- running the networks (which is what is on the cars). Right now NVIDIA with their SoCs is quite dominant but there are a number of custom ASIC solutions in the wings-- Google, for example, is using their own "TPU" in some applications. Here, I can imagine Qualcomm could be a major player.. their GPUs support OpenCL.. OpenCL is the alternative to CUDA but right now, at least, the support in OpenCV and most of the DL frameworks is not really up to grade.. while NVIDIA has done a lot of great work with CuDNN and RT-Tensor (GPU Inference Engine) ...

About

Avatar for hippy @hippy started