Accelerating artificial intelligence

Models for deep learning

Andrea Gillhuber, 31.08.2018, 07:00

Accelerating artificial intelligence

Mathworks announces that Matlab now offers integration with Nvidia TensorRT via GPU Coder. This will make it easier for engineers and scientists to develop new models for artificial intelligence and deep learning in Matlab.

Cuda code generation with NIVIDA TensorRT. © Mathworks

Matlab provides a complete workflow to quickly train, validate and deploy deep learning models. Engineers can utilize GPU resources without additional programming. With the new integration of NVIDIA TensorRT and GPU Coder, deep learning models developed in Matlab can run on NVIDIA GPUs with high throughput and low latency.

According to Mathworks, internal benchmarks show that CUDA code generated by Matlab in combination with TensorRT can provide Alexnet with five times higher performance and VGG-16 with 1.25 times higher performance than the deep learning inference of the corresponding nets in TensorFlow.

All benchmarks were run with Matlab R2018a with GPU Coder, TensorRT 3.0.1, TensorFlow 1.6.0, CUDA 9.0 and cuDNN 7 on an Nvidia Titan Xp GPU on a Linux-based 12-core Intel Xeon E5-1650 v3 PC with 64 GB RAM.

You might also be interested in

Deep Learning

Wenn die Maschine entscheidet

Beim Deep Learning lernt ein Computermodell, Klassifizierungsaufgaben direkt aus Bildern, Texten oder Tönen auszuführen – ohne Einflussnahme eines Menschen.

Maschinelles Lernen

Machine Learning implementieren

In diesem Jahr stand für MathWorks auf der Embedded World das Thema Implementierung von Algorithmen des Machine Learning und Deep Learning in Embedded Systemen im Fokus.

Automatica 2023

MVTec presents machine vision portfolio

MVTec Software is presenting its comprehensive machine vision portfolio at this year's automatica in Munich. From June 27 to 30, 2023, the company will be demonstrating the importance of machine vision as a key technology in automation.

Machine Learning

User-friendly and efficient image processing

There is no doubt that artificial intelligence offers enormous potential. However, many companies in the machine vision sector lack the necessary expertise and time to deal with the topic in depth.

automatica 2023

Artificial intelligence in industrial image processing

Recent advances in AI-based image analysis make machine vision accessible to companies of all sizes, even without technical experience or programming knowledge. This reduces the time required to set up inspection applications and increases the...

SMARC module

Deep learning inference and UHD media processing

With the Finlay computer-on-module, Seco presents a new embedded solution. It is based on the Intel Atom CPU of the x7000E series, Intel Core i3 CPU and Intel processors of the N series. Finlay has been specially developed for media processing.

Embedded Computing

The tiger among processors

Since the move towards the Industrial Internet of Things (IIoT) and Industry 4.0 has picked up speed, processor boards and modules must be able to process exponentially growing amounts of data.

Industry 4.0

From the digital platform to the digital ecosystem

Digital platforms are springing up like mushrooms. Every company involved in digitalization and Industry 4.0 has at least considered the potential benefits of its own digital platform, and many are already on the market. There are numerous...