摘要

本文课题构建一个深度学习算法实验平台，是为颈动脉医疗图像深度学习识别斑块提供服务的。制作一个集成的深度学习工具包，颈动脉医疗图像的深度学习代码都可以在本平台上运行，平台基于TensorFlow、Caffe、CNTK、Keras、Theano等多种深度学习框架并可以部署在同一硬件环境，因此可以测试颈动脉医疗图像深度学习在不同深度学习框架上的准确率和效率。

深度学习在语音、图像识别，自然语言处理等多个领域都取得了可观的成果，无论是大众和媒体的目光，或是产业界和资本，都在疯狂地追逐着“深度学习”这一概念。深度学习框架的不断涌现，深度学习的热潮正在掀起。

对于上述问题，该文尝试构建一个深度学习算法实验平台，旨在为深度学习用户提供一站式服务，为其在深度学习领域的探索降低门槛并提供流畅的体验。在现今软件配置和管理部署复杂的背景下，以简化应用部署，方便扩展，平台基于Docker容器技术，设计实现一种轻量级的应用平台。

关键词：Docker，深度学习，实验平台

ABSTRACT

This paper constructs a deep learning algorithm experimental platform, which is a service for the deep learning of plaques for carotid artery medical images. The deep learning kit for carotid medical image can be run on this platform. The platform is based on TensorFlow, Caffe, CNTK, Keras, Theano and other depth learning frameworks and can be deployed in the same hardware environment, so you can test the accuracy and efficiency of carotid medical image depth learning in different depth learning frameworks.

Deep learning in voice, image recognition, natural language processing and other fields have made considerable achievements, both the public and the media's eyes, or industry and capital, are crazy to chase the "deep learning" concept The Deep learning framework of the emerging, deep learning boom is set off.

For the above problem, the paper attempts to build a deep learning algorithm experimental platform designed to provide one-stop service for the deep learning users, to reduce the threshold and provide a smooth experience for its exploration in the deep learning. In the context of today's software configuration and management deployment, to simplify application deployment, easy to expand, the platform based on Docker container technology, design and implementation of a lightweight application platform.

Key words：Docker, deep learning, experiment platform

2.1.3 数据卷(Data Volume)