Infuse AI to Your Enterprise Yonghua LIN, IBM Research IBM - PowerPoint PPT Presentation

Infuse AI to Your Enterprise Yonghua LIN, IBM Research IBM Distinguished Engineer Leader of AI System Research

IBM Opens New Era of Artificial Intelligence Solution Supporting Technologies IBM Watson Cognitive Transportation Finance & Smarter City Medical Insurance Cognitive Media & Automobile Manufacture Retailer Entertainment AI Vision, Acoustic, Deep Learning & Machine AI Cloud Language, Conversation Learning Computing Watson won humans in Jeopardy In-memory Quantum Deep Learning Neural Chips Computing Computing Systems 2

Global research teams formed the base to drive AI innovation in IBM  IBM’s 2016 patent output features more than 2,700 patents for inventions related to AI, cognitive computing and cloud computing  IBM Research has deep understanding on vision technology and published hundreds of high quality papers on top AI and computer vision conferences , such as CVPR, AAAI, NIPS, etc.. 3

DEMO This is the first AI Highlight for TV entertainment show. Key words: Multi-model learning, AI for Media 4

AI-made Film / Cognitive Highlight To generate the highlight film for video will consume lots of time and human effort. • With AI technology, machine could understand each video frame and generate a • highlight film automatically. Multi-task models Generate index for video Scene detection Highlight strategies Action detection Inference on Emotion Create highlight New films video identification films by machine streams Face detection Audio identification One hour video: 80K entries and 100MB table Speech-to Text …

Challenges for enterprises going into AI TALENTS DATA 6

The hour glass for deep learning To build a team with deep learning expertise : 2 months ~ 1 year To prepare massive training data : ~ 10 man months To train a new model : 1 hour ~ weeks To give an AI inference result : < 1s

Steps for AI Deep Learning Development Usually, developers need following steps to develop a DNN model and make it usable for application Application development with inference API Start Package the new Configure the Define Prepare DL training DNN model Data Pre- DNN Model training DNN Model training training framework together with processing selection hyper- Training task Data preparation preprocessing into parameter inference proc. Most of enterprises are facing the challenges … • No experience on DNN design and develop • No experience on computer vision • No experience on how to build a platform to support enterprise scale deep learning, including data preparation, training, and inference

We can help deep learning for Vision easier – PowerAI Vision Deep knowledges of ML/DL and computer vision have been embedded into PowerAI Vision . User could use the deployed API for Start visual recognition Package the new Package the new Configure the Configure the Define Prepare DL training DL training DNN model DNN model Data Pre- Data Pre- DNN Model DNN Model training training DNN Model DNN Model training training framework framework together with together with processing processing selection selection hyper- hyper- Training Training preparation task Data preparation preprocessing into preprocessing into parameter parameter inference proc. inference proc. Steps automatically done by PowerAI Vision Easily develop AI application for Computer Vision !!

Example 1: AI for Product Quality Inspection (Manufacture) Inspect images of photoresist openings after having been exposed and developed ( 光刻是通过一系列生产步骤 , 将晶圆表面薄膜的特定部分去除 的工艺。被广泛用于集成电路的生产流程。 显影检查需要人工检验不合格的晶 圆 ，以便返工重新曝光、显影。 ) 显影检查：图形尺寸的偏差、光刻胶的污染、空洞、划伤，以及污点等。 Accuracy: 94.5% With PowerAI Vision , the manufacture could quickly build the auto defect inspection capability : Data import ：１５ min. • Data labeling ：５ min. • AI Vision training ：１０ min. •

PowerAI Vision : End-to-end DL Development and Operation for Computer Vision PowerAI Vision: Development Pipeline Data Training Inference Video and Data Custom Self-defined Inference API Inference Image Labeling Preprocessing Learning Training deployment for accelerator with visualized Cloud generation monitoring for edge Testing & Measurement ML & DL Libraries & Frameworks PowerAI Software Stacks Distributed Computing Data Lake & Data Stores PowerAI Accelerated Servers Storage

Custom Learning for Image and Video Custom learning for classification Custom learning for object detection Image and Video Analysis

Focus of PowerAI Vision AI for AI • •

Transfer Learning for Learning from Small Data Set • In lots of industry scenarios, we don’t have huge data set. • PowerAI Vision applied the optimized Transfer Learning technology for custom learning from small data set. Good base model Base models supported by PowerAI Vision Small data set from User 1. Select 3. One-click 2. Add user’s data scenario to start training data (base model) training Small data set, better accuracy, faster training 14

Data Augmentation for Learning from Small Data Set  Data Augmentation can enhance the classification accuracy and reduce overfitting for small datasets  Data Augmentation functions has been available on PowerAI Vision Fig.1 User could “one - click” and select different data Fig.2 Data augmentation could improve the augmentation methods accuracy significantly Medical image analysis for cerebral hemorrhage ( 脑出血 ) (Original data: 157 pic.) Accuracy: 97.9% 15

DL for DL: Learning to optimize parameters for visual analysis • Through machine learning, PowerAI Vision will automatically tune parameters to achieve good accuracy for different training cases defined by users. • In the following test case, our auto-tuning DL network could outperform the fix manual configuration (default) by >6% . And it could achieve the same accuracy (e.g. 90%) with much less training time (e.g. <1/3 ). Fig. 1 Performance comparison for object detection 18 parameters have been tuned, including • Caffe training parameters • Neural network parameters • Object detection parameters. Test data set: object detection for helmet and safety vest Auto-tune 16

Semi auto-labeling : Reduce the time for data annotation  Semi – auto labeling : To use AI technology for releasing most of human work for labeling ( 10x ~ 50x ) System will Auto – labeling Manually label Human review learn the objects small data set by machines and adjust for labeling 17

DEMO This is a demo for retail scenario. In this demo, audience can see how to use PowerAI Vision to learn and detect shopping cart easily. Key words: custom learning for object detection, auto-labeling 18

PowerAI Vision for Retail Business 1. Real-time understand your customer and make 2. Client flow analysis recommendation • • Hot area analysis Age and gender • • Client behavior What she wears • Styles: office lady, analysis fashion, etc. • Queue length • Shopping with kids, monitoring friend, or parents • … 3. Shelves management 4. Risk detection for security 5. Anti-Lost • Fighting • Goods • Goods detection counting on tracking at the • Wearing mask the shelves counter • Goods • illegal detection • Occupation in distribution on behavior sensitive area the shelves analysis (e.g. exit path)

企业级的人工智能系统需要极高的综合技术能力：车载辅助驾驶 系统 汽车企业需要构建端到端的深度学习开发平台，支持从数据中心到车载系统的 AI 研发 实时低时延：每帧 9ms 路测效果 小目标 超小目标检测 : 12x12 像素 多种天气及光照条件 单目高精准测距 PowerAI Vision 深度学习开发 FPGA 嵌入式系统 FPGA (Znyq70xx series) 训练数据集 一键启动深 监测训练过程 自动化生成 FPGA 标注视频训 预处理 度学习训练 练数据 加速器软件 20

PowerAI Vision : Infuse AI into Your Enterprise TALENTS DATA Innovation with AI for AI 21