Deep Learning in MATLAB

Transcription

1 Deep Learning in MATLAB 성호현부장 2015 The MathWorks, Inc. 1

2 Deep Learning beats Go champion! 2

3 AI, Machine Learning, and Deep Learning Artificial Intelligence Any technique that enables machines to mimic human intelligence Machine Learning Statistical methods enable machines to learn tasks from data without explicitly programming Deep Learning Neural networks with many layers that learn representations and tasks directly from data 1950s 1980s 2015 FLOPS Thousand Million Quadrillion 3

4 What is can Deep Learning do for us? (An example) 4

5 Example 1: Object recognition using deep learning 5

6 Object recognition using deep learning Training (GPU) Prediction Millions of images from 1000 different categories Real-time object recognition using a webcam connected to a laptop 6

7 What is is Machine Deep Learning? 7

8 Machine Learning vs Deep Learning Machine Learning We specify the nature of the features we want to extract and the type of model we want to build. 8

9 Machine Learning vs Deep Learning Deep Learning We need only specify the architecture of the model 9

10 Deep learning is a type of machine learning in which a model learns to perform tasks like classification directly from images, texts, or signals. Deep learning performs end-to-end learning, and is usually implemented using a neural network architecture. Deep learning algorithms also scale with data traditional machine learning saturates. 10

11 Why is Deep Learning So Popular Now? AlexNet Human Accuracy Source: ILSVRC Top-5 Error on ImageNet 11

12 Two Approaches for Deep Learning 1. Train a Deep Neural Network from Scratch 2. Fine-tune a pre-trained model (transfer learning) 12

13 Pains In Deep Learning Expertise Time to Train Data 13

14 Example: Vehicle recognition using deep transfer learning Cars Trucks SUVs Big Trucks Vans 5 Category Classifier 14

15 Import the Latest Models for Transfer Learning Pretrained Models* AlexNet VGG-16 VGG-19 GoogLeNet Inception-v3 ResNet50 ResNet-101 Inception-resnet-v2 SqueezeNet MobileNet(coming soon) Import Models from Frameworks Caffe Model Importer TensorFlow-Keras Model Importer Onnx - Importer/ Exporter (Coming Soon) AlexNet PRETRAINED MODEL Caffe I M P O R T E R VGG-16 PRETRAINED MODEL GoogLeNet PRETRAINED MODEL ResNet-50 PRETRAINED MODEL TensorFlow- Keras I M P O R T E R ResNet-101 PRETRAINED MODEL Inception-v3 M O D E L S * single line of code to access model 15

16 Detection and localization using deep learning Regions with Convolutional Neural Network Features (R-CNN) 16

17 What is semantic segmentation? 17

18 Localization using deep learning Original Image ROI detection Pixel classification 18

19 Semantic Segmentation Network Boat Airplane Other classes 19

20 Semantic Segmentation Network 20

21 Semantic Segmentation Demo CamVid Dataset 1. Segmentation and Recognition Using Structure from Motion Point Clouds, ECCV Semantic Object Classes in Video: A High-Definition Ground Truth Database,Pattern Recognition Letters 21

22 Semantic Segmentation CamVid Dataset 1. Segmentation and Recognition Using Structure from Motion Point Clouds, ECCV Semantic Object Classes in Video: A High-Definition Ground Truth Database,Pattern Recognition Letters 22

23 I love to label and preprocess my data ~ Said no engineer, ever. 23

24 Ground truth Labeling How do I label my data? New App for Ground Truth Labeling Label pixels and regions for semantic segmentation Data 24

25 Attributes and Sublabels NEW in 25

26 Types of Datasets Numeric Data Time Series/ Text Data Image Data ML or LSTM LSTM or CNN CNN 26

27 Analyzing signal data using deep learning Signal Classification using LSTMs Speech Recognition using CNNs 27

28 Deep learning features overview Classification Regression Semantic segmentation Object detection Scalability Multiple GPUs Cluster or cloud Custom network layers Import models Caffe Keras/TensorFlow Data augmentation Hyperparameter tuning Bayesian optimization Python MATLAB interface LSTM networks Time series, signals, audio Custom labeling API for ground-truth labeling automation Superpixels Data validation Training and testing 28

29 Prediction Performance: Fast with GPU Coder Images/Sec Why is GPU Coder so fast? Analyzes and optimizes network architecture Invested 15 years in code generation TensorFlow MATLAB MXNet GPU Coder AlexNet ResNet-50 VGG-16 Using CUDA v9 and cudnn v7 29

30 Overview of deep learning deployment options How do I deploy my model? Create Desktop Apps GPU Coder Introducing: GPU Coder- Convert to NVIDIA CUDA code Run Enterprise Solution Generate C and C ++ Code Deploy / Share Target GPUs Generate C and C ++ Code 30

31 GPU Coder Fills a Gap in Our Deep Learning Solution Training Inference Access Data Preprocess Select Network Train Deploy Image Acq. Image Processing Neural Network PCT GPU Coder Computer Vision 31

32 Deploying to CPUs Intel MKL-DNN Library Deep Learning Networks GPU Coder NVIDIA TensorRT & cudnn Libraries ARM Compute Library 32

33 MATLAB products for deep learning Required products Neural Network Toolbox Parallel Computing Toolbox Image Processing Toolbox Computer Vision System Toolbox Recommended products Statistics and Machine Learning Toolbox MATLAB Coder GPU Coder Automated Driving System Toolbox 33

34 Deep learning features overview Classification Regression * Semantic segmentation Object detection * Scalability * Multiple GPUs Cluster or cloud Custom network layers * Import models * Caffe Keras/TensorFlow Data augmentation * Hyperparameter tuning * Bayesian optimization Python MATLAB interface * LSTM networks * Time series, signals, audio Custom labeling * API for ground-truth labeling automation Superpixels Data validation * Training and testing * We can cover in more detail outside this presentation 34

35 Thank you! Deep Learning Onramp MATLAB 35

36 Deep learning in automated driving 36

37 Deep Learning Onramp Get started using deep learning methods to perform image recognition. Free access for everyone Interactive exercises and short video demonstrations Work on real-life image recognition problems Topics include: Convolutional neural networks Working with pre-trained networks Transfer learning Evaluating network performance 37

38 Convolutional Neural Networks (CNN) Edges Shapes Objects 38

39 Deep Reinforcement Learning (E.g. Deep Q Network) Policy is a sequence of actions to observations to get maximum reward AGENT POLICY Reinforcement Learning finds the optimal policy maximizing the reward Reinforcement Learning adapts to changes in environment by improving the policy OBSERVATION S REWARD ACTIONS No need for explicit model (model-free) 39

40 Google Deepmind s Deep Q Learning playing Atari Breakout 40

41 41