您现在的位置:
产品服务

深度学习

0.00
0.00
适用平台:
  

简介:

深度学习是人工智能(Artificial Intelligence)和机器学习(Machine Learning)领域中的一个分支,通过建立类似于人脑的多层神经网络,来实现目标检测、语言识别、语言翻译等任务,且具有非常高的准确率。卷积神经网络(CNNConvolutional Neural Network)和循环神经网络(RNN, Recurrent Neural Network)是深度学习中两种常见的网络模型。可提供针对多种FPGA加速板卡的CNN/RNN IP库以及基于YOLOVGGResnet等算法的整体解决方案。

 

特点:

提供Turnkey解决方案,包括用户API、驱动、FPGA设计方案以及加速板卡;

支持从caffe等深度学习框架中提取已训练好的模型,自动解析网络参数和配置FPGA运行环境。

支持定制化网络模型的设计,支持卷积层/池化层/归一化层及其参数的动态配置

GPU相比,FPGA加速方案具有较高的效能功耗比。

 

 

Overview

Deep learning is a subset of AI and machine learning that uses multi-layered artificial neural networks to deliver state-of-the-art accuracy in tasks such as object detection, speech recognition, language translation and others. Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) are two types of popular network models. We provide CNN/RNN IP libraries targeted at multiple FPGA acceleration boards, and integrated solutions based on the algorithms of YOLO, VGG and Resnet.

Features:

Provides turnkey solutions, including user API, drivers, FPGA logic design and acceleration boards;

Supports direct use of trained models extracted from caffe deep-learning framework, automatically paring network parameters, and configuring FPGA runtime environment;

Provides design service for customized network models, and supports the runtime configuration in term of parameters of Convolution/Pooling/Normalization layers.

Performance per watt of FPGA accelerating solutions is better than that of GPU