NVIDIA Reveals 3 Big Updates To Deep-Learning Software Platform
Three big updates are coming to NVIDIA’s deep-learning software platform, going by a recent announcement from the company.
These updates — NVIDIA DIGITS 4, CUDA Deep Neural Network Library (cuDNN) 5.1, and the new GPU Inference Engine (GIE) — include DIGITS 4’s new object detection workflow, one which will allow data researchers to train deep neural networks to identify and locate traffic signs, pedestrians, faces, vehicles, etc, in the mass of images.
This workflow can in turn be used for advanced autonomous driving systems, security and surveillance work, satellite imagery object identification, medical diagnostics, etc.
Of note, deep-neural network training typically involves repeated tuning of parameters by researchers in order to get high accuracy from the model, while DIGITS 4 can apparently automatically train neural networks, across a range of different tuning parameters, thereby notably lessening the time required to reach the “most” accurate solution.
The DIGITS 4 release candidate will be available as a free download this week for members of the NVIDIA developer program, reportedly.
Here’s an overview of the other two updates (cuDNN and GIE), drawing from the press release:
- NVIDIA cuDNN provides high-performance building blocks for deep learning used by all leading deep learning frameworks. Version 5.1 delivers accelerated training of deep neural networks, like University of Oxford’s VGG and Microsoft’s ResNet, which won the 2016 ImageNet challenge.
- Each new version of cuDNN has delivered performance improvements over the previous version, accelerating the latest advances in deep learning neural networks and machine learning algorithms.
- The GPU Inference Engine (GIE) is a high-performance deep learning inference solution for production environments. GIE optimizes trained deep neural networks for efficient runtime performance, delivering up to 16x better performance per watt on an NVIDIA Tesla M4 GPU vs the CPU-only systems commonly used for inference today.
- Using GIE, cloud service providers can more efficiently process images, video and other data in their hyperscale data center production environments with high throughput. Automotive manufacturers and embedded solutions providers can deploy powerful neural network models with high performance in their low-power platforms.
As noted previously, NVIDIA supplies components for Tesla Motors vehicles. How much it is working with Tesla on deep learning is unclear, though.
Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.
CleanTechnica Holiday Wish Book
Our Latest EVObsession Video
CleanTechnica uses affiliate links. See our policy here.
Keep an eye, actually both eyes on Nvidia. They are a perfect company for that application and they are positioning themselves to be one of the market leaders.
But is there a source for Nvidia working with Tesla on deep learning and autonomous driving? I only know that Nvidia provides Tesla with the Tegra Chip to power the car-computer.
Why would you keep more eyes on Nvidia than on their competitors?
AMD’s GPUs and the coprocessors of Intel, IBM and others are equally capable, and they use the open, portable OpenCL standard rather than CUDA. A wide range of software packages and programming environments (MATLAB, R, most molecular dynamics software etc) also use their own HPC/GPU-accelerated frameworks.
Tesla Motors also doesn’t specifically demand CUDA in its job opening (i.e. https://www.teslamotors.com/careers/job/computer-visionscientist-29346 ). MATLAB OpenGL/CL are mentioned alongside CUDA.
This article reads like paid content to me, even though it not marked as such.
Who said anything about ‘more than other competitors’? Don’t twist my words.
And I never heard of AMD or Intel working on deep learning for autonomous cars.
The Intel Scalable System Framework (with the Phi coprocessor) is widely used for autonomous vehicles, and it owns Itseez which develops software for autonomous vehicles.
AMD does not have dedicated autonomous vehicle software, but its hardware is of course widely used and it is an active contributor to the OpenCL framework (as well as dedicated deep learning frameworks like Caffe).
IBM has the Watson platform that is used for automative applications, and is of course a major seller of HPC hardware in general. Its framework is arguably more advanced than Nvidia’s, at least on the software side.
Also note the article only tangentially mentions autonomous driving. The software mentioned is general purpose HPC software and not specific to autonomous vehicles.
I’m not trying to dismiss Nvidia, but I’m a little disturbed by this article. Either it is a paid press release, in which case that should at least be mentioned in the title, or the author didn’t bother to even look up any competing products. Both possibilities are worrying; CT is normally relatively objective.