第 1 章 深度學習簡介 ············································································ 1
1.1 深度學習的發展 ·······································································1
1.2 深度學習的應用及研究方嚮 ···················································3
1.3 深度學習工具介紹和對比 ·······················································4
1.3.1 Caffe·················································································4
1.3.2 TensorFlow ······································································5
1.3.3 Torch ················································································6
1.4 小結 ···························································································7
第 2 章 深度學習基本理論 ····································································9
2.1 深度學習的基本概念 ·······························································9
2.2 深度學習的訓練過程 ·····························································13
2.3 深度學習的常用模型和方法 ·················································14
2.4 小結 ·························································································20
第 3 章 深度學習環境搭建 ································································· 23
3.1 Caffe 安裝 ···············································································23
3.1.1 安裝 Caffe 的相關依賴項·············································24
3.1.2 安裝 NVIDIA 驅動 ·······················································24
3.1.3 安裝 CUDA ···································································27
3.1.4 配置 cuDNN ··································································30
3.1.5 源代碼編譯安裝 OpenCV ············································32
3.1.6 編譯 Caffe，並配置 Python 接口 ································34
3.2 Caffe 框架下的 MNIST 數字識彆問題···································41
3.3 TensorFlow 安裝 ······································································42
3.3.1 基於 pip 安裝·································································42
3.3.2 基於 Anaconda 安裝 ······················································46
3.3.3 基於源代碼安裝····························································51
3.3.4 常見安裝問題································································56
3.4 TensorFlow 框架下的 CIFAR 圖像識彆問題·························59
3.5 Torch 安裝 ···············································································61
3.5.1 無 CUDA 的 Torch 7 安裝 ·············································61
3.5.2 CUDA 的 Torch 7 安裝 ··················································61
3.6 Torch 框架下 neural-style 圖像閤成問題······························62
3.7 小結 ·························································································74
第 4 章 人臉識彆 ················································································· 75
4.1 人臉識彆概述 ·········································································75
4.2 人臉識彆係統設計 ·································································76
4.2.1 需求分析········································································76
4.2.2 功能設計········································································77
4.2.3 模塊設計········································································78
4.3 係統生産環境部署及驗證 ·····················································81
4.3.1 抽幀環境部署································································81
4.3.2 抽幀功能驗證································································82
4.3.3 OpenFace 環境部署·······················································82
4.3.4 OpenFace 環境驗證·······················································84
4.4 批量生産 ·················································································90
4.5 小結 ·······················································································102
第 5 章 車輛識彆 ···············································································103
5.1 概述 ·······················································································103
5.2 係統設計 ···············································································104
5.2.1 需求分析······································································104
5.2.2 功能設計······································································104
5.2.3 模塊設計······································································105
5.3 係統生産環境部署及驗證 ···················································106
5.3.1 生産環境部署······························································106
5.3.2 項目部署······································································107
5.3.3 環境驗證······································································108
5.4 批量生産 ···············································································109
5.5 小結 ·······················································································117
第 6 章 不良視頻識彆 ······································································· 119
6.1 概述 ·······················································································119
6.2 不良圖片模型簡介 ·······························································120
6.3 係統設計 ···············································································122
6.4 係統部署及係統測試驗證 ···················································123
6.5 批量生産 ···············································································125
6.5.1 批量節目元數據信息檢索與篩選······························125
6.5.2 基於 FFmpeg 的 SDK 抽取視頻 I 幀 ··························126
6.5.3 基於膚色比例檢測的快速篩查··································128
6.5.4 基於 Caffe 框架的不良圖片檢測································128
6.6 小結 ·······················································································129
第 7 章 集群部署與運營維護 ··························································· 131
7.1 認識 Docker···········································································131
7.2 基於 Docker 的 TensorFlow 實驗環境··································134
7.3 運營維護 ···············································································137
7.4 小結 ·······················································································138
參考文獻································································································139
· · · · · · (收起)