人工智能

对于人工智能这艘即将腾飞的火箭,深度学习是引擎、大数据是燃料。这里列举了供应商提供的云服务或者技术,以及相应的论文方便查询。

服务

机器学习供应商包括微软、亚马逊、谷歌以及阿帕奇基金会:

Microsoft Amazon Google Apache
机器学习 Azure ML Amazon ML Cloud ML  Spark MLlib
深度学习 CNTK MXNet TensorFlow MXNet
SINGA
认知服务 Language API
Speech
Vision
Knowledge
Amazon Lex
Amazon Polly
Amazon Rekognition
Natural Language API
Cloud Speech API
Cloud Vision API
Cloud Translation API

教程

  • 斯坦福公开课Machine Learning,Andrew Ng深入浅出的讲解有口皆碑。
  • Deep Learning for Computer Vision with Python,作者Adrian Rosebrock是计算机视觉和机器学习的PhD,维护PyImageSearch.com的同时在社区也是非常积极。传统的计算机视觉教程往往针对数据科学家讲解算法,而这本书针对数据工程师注重实践,通过数据集和代码帮助计算机视觉的爱好者快速掌握基本的工程开发能力。

论文

以下收集了部分电子书以及论文方便查询技术细节。

深度框架

  • TensorFlow – Large-scale Machine Learning on Heterogeneous Systems
  • Caffe – Convolutional Architecture for Fast Feature Embedding
  • Chainer – a Next-Generation Open Source Framework for Deep Learning
  • MXNet – A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
  • TensorLayer – A Versatile Library for Efficient Deep Learning Development
  • SINGA – A Distributed Deep Learning Platform

训练数据

  • Automated Flower Classification over a Large Number of Classes
  • ImageNet – A Large-Scale Hierarchical Image Database
  • Microsoft COCO – Common Objects in Context

工程优化

  • In-Datacenter Performance Analysis of a Tensor Processing Unit
  • Large Scale Distributed Deep Networks
  • Scalable Distributed DNN Training Using Commodity GPU Cloud Computing

卷积网络

  • Handwritten Digit Recognition with a Back-Propagation Network (LeNet)
  • ImageNet Classification with Deep Convolutional Neural Networks (AlexNet)
  • Network in Network (NIN)
  • Very Deep Convolutional Networks for Large-Scale Image Recognition (VGGNet)
  • Going Deeper with Convolutions (Inception V1)
  • Batch Normalization – Accelerating Deep Network Training by Reducing Internal Covariate Shift (Inception V2)
  • Rethinking the Inception Architecture for Computer Vision (Inception V3)
  • Inception-v4 – Inception-ResNet and the Impact of Residual Connections on Learning (Inception V4)
  • Deep Residual Learning for Image Recognition (ResNet V1)
  • Identity Mappings in Deep Residual Networks (ResNet V2)
  • Aggregated Residual Transformations for Deep Neural Networks (ResNeXt)
  • MobileNets – Efficient Convolutional Neural Networks for Mobile Vision Applications
  • Visualizing and Understanding Convolutional Networks
  • Understanding Deep Image Representations by Inverting Them
  • Understanding Neural Networks Through Deep Visualization

图像处理

  • A Neural Algorithm of Artistic Style
  • Image Style Transfer Using Convolutional Neural Networks
  • A Learned Representation for Artistic Style
  • Image Super-Resolution using Deep Convolutional Networks
  • Accurate Image Super-Resolution Using Very Deep Convolutional Network
  • Perceptual Losses for Real-Time Style Transfer and Super-Resolution
  • Neural Style Transfer – A Review
  • Deep Colorization
  • Colorful Image Colorization

文字识别

  • CAPTCHA Recognition with Active Deep Learning
  • Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks
  • Neural Network CAPTCHA Crackers

安防监控

  • The German Traffic Sign Recognition Benchmark: A multi-class classification competition
  • Traffic Sign Recognition with Multi-Scale Convolutional Networks
  • A Large-scale Car Dataset for Fine-Grained Categorization and Verification
  • 3D Object Representations for Fine-Grained Categorization
  • Monza – Image Classification of Vehicle Make and Model Using Convolutional Neural Networks and Transfer Learning

精准医疗

  • A Deep Learning Architecture for Image Representation, Visual Interpretability and Automated Basal-Cell Carcinoma Cancer Detection
  • Deep Learning for Identifying Metastatic Breast Cancer
  • Detecting Cancer Metastases on Gigapixel Pathology Images
  • Dermatologist-level Classification of Skin Cancer with Deep Neural Networks