Intel provides access to clDNN [high-performance library for deep learning]
The official project repository was launched just a couple of days ago. We will tell you a little more about this news and provide useful sources on the topic.
/ Flickr / fdecomite / CC
The main area of ​​application of clDNN (Compute Library for Deep Neural Networks) is image recognition and segmentation. High-performance library works with DL-applications that involve the acceleration of the construction of logical conclusions (inference) using Intel HD / Iris Graphics.
')
clDNN was released under the Apache 2.0 license and supports AlexNet , VGG, GoogleNet, and ResNet (a little about the last three). The library consists of blocks optimized for working with convolutional neural networks in C and C ++.
Deep neural networks (DNN) are also used for video analysis, natural language processing (NLP), and problem solving in areas such as autonomous driving. DNN is very demanding in terms of computational resources and the efficiency of certain implementations of algorithms, therefore, new findings in this area are always met with great interest. For example, the company recently introduced the free Intel Deep-Learning Deployment Toolkit , which is part of a thematic SDK for working with computer vision technologies.
“01” is the open source Intel community. The company's engineers are actively involved in its work and replenish the collection of open solutions. Each project has its own microsite (a couple of examples: OpenStack , Crosswalk , Developer Journey ).
PS We in 1cloud are developing a virtual infrastructure provider. A detailed account of this work is here and here . A little bit about our API is here .
What else do we write on Habré:
/ Flickr / fdecomite / CC
The main area of ​​application of clDNN (Compute Library for Deep Neural Networks) is image recognition and segmentation. High-performance library works with DL-applications that involve the acceleration of the construction of logical conclusions (inference) using Intel HD / Iris Graphics.
')
clDNN was released under the Apache 2.0 license and supports AlexNet , VGG, GoogleNet, and ResNet (a little about the last three). The library consists of blocks optimized for working with convolutional neural networks in C and C ++.
Deep neural networks (DNN) are also used for video analysis, natural language processing (NLP), and problem solving in areas such as autonomous driving. DNN is very demanding in terms of computational resources and the efficiency of certain implementations of algorithms, therefore, new findings in this area are always met with great interest. For example, the company recently introduced the free Intel Deep-Learning Deployment Toolkit , which is part of a thematic SDK for working with computer vision technologies.
About the project "01"
“01” is the open source Intel community. The company's engineers are actively involved in its work and replenish the collection of open solutions. Each project has its own microsite (a couple of examples: OpenStack , Crosswalk , Developer Journey ).
PS We in 1cloud are developing a virtual infrastructure provider. A detailed account of this work is here and here . A little bit about our API is here .
What else do we write on Habré:
Source: https://habr.com/ru/post/329474/
All Articles