O'reilly AI Conference

Transcript

1. San Francisco, CA 17 - 20 September 2017 Bramantyo Adrian
2. What “It is an AI Event that focus on real-world implementations.” GDP Labs Confidential
3. PUT AI TO WORK
4. 1. Building Unbiased AI 2. Accelerating AI 3. Active Learning and Transfer Learning 4. Object Detection GDP Labs Confidential
5. Introduction + Internet Company IS NOT GDP Labs Confidential
6. Introduction + AI Company IS NOT GDP Labs Confidential
7. 1. Building Unbiased AI 2. Accelerating AI 3. Active Learning and Transfer Learning 4. Object Detection GDP Labs Confidential
8. Diversity Crisis in AI “We’ve already seeing society’s racial and gender biases being encoded into software that uses AI when built by such a homogeneous group.” Building Unbiased AI GDP Labs Confidential
9. Building Unbiased AI GDP Labs Confidential
10. Building Unbiased AI GDP Labs Confidential
11. Goals Building Unbiased AI GDP Labs Confidential
12. Building Unbiased AI GDP Labs Confidential
13. Ignoring - Failing to max a product to different groups - Failing to attract potential users Building Unbiased AI GDP Labs Confidential
14. 1. Building Unbiased AI 2. Accelerating AI 3. Active Learning and Transfer Learning 4. Object Detection GDP Labs Confidential
15. Accelerating AI GDP Labs Confidential
16. Manufacturing Automotive Healthcare Financial Transforming Many Industries Accelerating AI GDP Labs Confidential
17. CPU GPU FPGA ASIC Computational Substrate Accelerating AI GDP Labs Confidential
18. why Deep Learning ? Accelerating AI GDP Labs Confidential
19. Latest Model Not Always Relevant Model Accelerating AI GDP Labs Confidential
20. SSD model designed and optimized for PASCALVOC/MSCOCO dataset Know Your Model Provenance Accelerating AI GDP Labs Confidential
21. Smart hospital will generate over 3,000 GB PER DAY Self driving cars will generate over 4,000 GB PER DAY … EACH The average internet user will generate ~1.5 GB OF TRAFFIC PER DAY A connected plane will generate over 40,000 GB PER DAY A connected factory will generate over 1,000,000 GB PER DAY Invest in Data Manual annotations Training data Accelerating AI GDP Labs Confidential
22. 1. Building Unbiased AI 2. Accelerating AI 3. Active Learning and Transfer Learning 4. Object Detection GDP Labs Confidential
23. Active and Transfer Learning Active Attain good learning performance without demanding too many labeled data GDP Labs Confidential
24. Motivation Active and Transfer Learning GDP Labs Confidential
25. Human Active AI Classifier Output Human Annotation Confident N ot Confident Active Learning Active and Transfer Learning GDP Labs Confidential
26. Transfer It’s very expensive to train deep learning Active and Transfer Learning GDP Labs Confidential
27. Transfer as feature extractor Active and Transfer Learning GDP Labs Confidential
28. Transfer as fine tuning Active and Transfer Learning GDP Labs Confidential
29. 1. Building Unbiased AI 2. Accelerating AI 3. Active Learning and Transfer Learning 4. Object Detection GDP Labs Confidential
30. Computer Vision Object Detection GDP Labs Confidential
31. Traditional Machine Learning to Deep Learning Object Detection Feature Engineering challenges: - Human Scaling - Computational Scaling GDP Labs Confidential
32. Object Detection Feature Learning challenges: - Need large datasets - Model become complex Traditional Machine Learning to Deep Learning GDP Labs Confidential
33. Object Detection CIFAR, ImageNet PascalVOC MSCOCO KITTI GDP Labs Confidential
34. Convolutional Neural Network Object Detection GDP Labs Confidential
35. Object Detection Object Detection GDP Labs Confidential
36. Object Detection History Object Detection Region Proposed Approach Direct Classification Refined Classification GDP Labs Confidential
37. Region Proposed Approach Object Detection - Run through image to detect regions - Process regions extensively: - Bounding boxes - Predict class GDP Labs Confidential
38. Direct Classification Object Detection - BB and class are predicted directly with a single network GDP Labs Confidential
39. Refined Classification Object Detection - Add another branch that predict pixel-wise classes GDP Labs Confidential
40. Training Vision Model Challenges Object Detection GDP Labs Confidential
41. Training Vision Model Challenges Object Detection Batch size epochs Top-1 Accuracy Hardware Cost ($) Time 256 90 73.0% 1 DGX station 129,000 21h 8192 90 72.7% 1 DGX station 129,000 21h 8192 90 72.7% 32 DGX station 4.1 million 1h 32K 90 72.4% 512 KNLs 1.2 million 1h GDP Labs Confidential
42. Object Detection
43. THANK YOU