Deep Learning e Visão Computacional com CNTK
0 likes44 views
This document provides an overview of deep learning and computer vision. It discusses how vision is an ambiguous problem affected by small variations, noise, and other factors. It then shows how teaching a computer what a cat is can be challenging. The document outlines areas of machine learning like computer vision, sound analytics, and language analytics. It also discusses concepts like vectors, matrices, tensors, perceptrons, and generative adversarial networks. Finally, it discusses how Microsoft is investing in technologies like FPGAs, CPUs, and GPUs to enable AI training at large scale in the cloud.
![[8]](https://image.slidesharecdn.com/vssummit2018-180906164754/85/Deep-Learning-e-Visao-Computacional-com-CNTK-19-320.jpg)
![[1,2,3]](https://image.slidesharecdn.com/vssummit2018-180906164754/85/Deep-Learning-e-Visao-Computacional-com-CNTK-20-320.jpg)
![[1,2,3]
[4,5,6]
[7,8,9]](https://image.slidesharecdn.com/vssummit2018-180906164754/85/Deep-Learning-e-Visao-Computacional-com-CNTK-21-320.jpg)
Deep Learning e Visão Computacional com CNTK
- 1. VITOR MERIAT @VITORMERIAT Deep Learning e Visão Computacional com CNTK #VSSUMMIT
- 3. Agenda
- 5. Visão é um problema inverso e mal-posto • Pequenas variações na imagem levam a interpretações distintas; • Visão é ambígua • Ruído piora o problema • Ambiguidades • Distorções • Paradoxos • Ficções
- 7. Eu quero ensiar para o meu computador que isso é um gato O que é um gato? Eu sou um gato
- 8. Vamos ensinar esse computar
- 9. Isso é um gato. Agora eu já sei.
- 10. WTF?!?!
- 11. WTF?!?!
- 15. Área da Computação que estuda e desenvolve algoritmos que “tentam” imitar o processo da Inteligência Humana, de aprender, sentir, identificar e resolver problemas. Artificial Taste & Smell (Sensors) Computer Vision Sound Analytics Language AnalyticsArtificial Touch (Robots) Machine Learning Speech Analytics Language Analytics https://goo.gl/eN6zvh
- 19. [8]
- 20. [1,2,3]
- 24. Perceptron
- 25. What?
- 29. A faster, more efficient, more intelligent cloud Data explosion: 2013 4.4 ZB - 2020 44 ZB ML, DNN, AI are driving requirements up faster Autonomous decision making Real-time insights into connected devices Interactive user experiences Cloud-scale services Searches and recommendations (Indexing the Internet!) The need for SCALE The need for LOW-LATENCY The need for THROUGHPUT 1 0 0 1 1 0 1 0 1 0 2013 1 0 0 1 1 0 1 0 1 0 2020 4.4 ZB 44 ZB 0 1 0 0 0 1 0 1 0 1 1 1 0 1 1 0 1 0 1 0 1 0 1 1 0 0 0 1 1 0 0 1 0 0 0 1 0 1 0 1 1 0 0 1 1 0 1 0 1 0 0 0 1 0 0 1 1 0 1 1 1 0 1 1 1 0 1 0 1 0 0 0 0 1 0 0 1 1 1 0 0 1 1 0 0 0 1 0 1 1 Source: IDC 2014
- 31. Azure Batch AI Training AI Training At Scale with Azure
- 32. Autodesk 3ds Max / Maya VM Renderer VM Renderer VM Renderer Integrated Client Plugin Azure Batch Monitoring Reporting Single bill
- 33. FPGAs EVALUATION CPUs and FPGAs, ASICs under investigation EFFICIENCY TRAINING CPUs and GPUs, limited FPGAs, ASICs under investigation Control Unit (CU) Registers Arithmetic Logic Unit (ALU) + + + + + + + FLEXIBILITY CPUs GPUs ASICs
- 34. Microsoft has the world’s largest cloud investment in FPGAs Multiple Exa-Ops of aggregate AI capacity We have built powerful DNN serving platform on our FPGA fabric FPGAs ideal for adapting to rapidly evolving ML CNNs, LSTMs, MLPs, reinforcement learning, feature extraction, decision trees, etc. Inference-optimized numerical precision Custom binarized, ternarized, tiny precision nets Sparsity, deep compression for larger, faster models Tens to hundreds of TOPS of effective inference throughput at low batch sizes Ultra-low latency serving on modern DNNs >10X better than CPUs and GPUs Scale to many FPGAs in single DNN service Performance Flexibility Scale