imagenet
play

Imagenet Xavier Gir-i-Nieto ImageNet ILSRVC Li Fei-Fei, How were - PowerPoint PPT Presentation

Jun 15, 2023 •316 likes •855 views

Day 2 Lecture 4 Imagenet Xavier Gir-i-Nieto ImageNet ILSRVC Li Fei-Fei, How were teaching computers to understand pictures TEDTalks 2014. Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L.

  1. Day 2 Lecture 4 Imagenet Xavier Giró-i-Nieto

  2. ImageNet ILSRVC Li Fei-Fei, “How we’re teaching computers to understand pictures” TEDTalks 2014. Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L. (2015). Imagenet large scale visual 2 recognition challenge. arXiv preprint arXiv:1409.0575 . [web]

  3. ImageNet ILSRVC Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L. (2015). Imagenet large scale visual recognition challenge. arXiv preprint arXiv:1409.0575 . [web] 3

  4. ImageNet ILSRVC ● 1,000 object classes (categories). ● Images: ○ 1.2 M train ○ 100k test. 4

  5. ImageNet ILSRVC ● Top 5 error rate Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L. (2015). Imagenet large scale visual recognition challenge. arXiv preprint arXiv:1409.0575 . [web]

  6. ImageNet ILSRVC Image Classification 2012 Based on SIFT + Fisher Vectors Slide credit: -9.8% Rob Fergus (NYU) Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L. (2014). Imagenet large scale visual recognition challenge. arXiv 6 preprint arXiv:1409.0575 . [web]

  7. AlexNet (Supervision) Orange A Krizhevsky, I Sutskever, GE Hinton “Imagenet classification with deep convolutional neural networks” Part of: Advances in Neural Information Processing Systems 25 (NIPS 2012) Slide credit: Junting Pan, “Visual Saliency Prediction using Deep Learning Techniques” (ETSETB-UPC 2015) 7

  8. AlexNet (Supervision) Slide credit: Junting Pan, “Visual Saliency Prediction using Deep Learning Techniques” (ETSETB-UPC 2015) 8

  9. AlexNet (Supervision) Slide credit: Junting Pan, “Visual Saliency Prediction using Deep Learning Techniques” (ETSETB-UPC 2015) 9

  10. AlexNet (Supervision) 10 Image credit: Deep learning Tutorial (Stanford University)

  11. AlexNet (Supervision) 11 Image credit: Deep learning Tutorial (Stanford University)

  12. AlexNet (Supervision) 12 Image credit: Deep learning Tutorial (Stanford University)

  13. AlexNet (Supervision) f(x) = max(0,x) Rectified Linear Unit (non-linearity) Slide credit: Junting Pan, “Visual Saliency Prediction using Deep Learning Techniques” (ETSETB-UPC 2015) 13

  14. AlexNet (Supervision) Dot Product Slide credit: Junting Pan, “Visual Saliency Prediction using Deep Learning Techniques” (ETSETB-UPC 2015) 14

  15. ImageNet ILSRVC ImageNet Classification 2013 Slide credit: Rob Fergus (NYU) Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L. (2015). Imagenet large scale visual recognition challenge. arXiv 15 preprint arXiv:1409.0575 . [web]

  16. Zeiler-Fergus (ZF) The development of better convnets is reduced to trial-and- Visualization can help in error. proposing better architectures. Zeiler, M. D., & Fergus, R. (2014). Visualizing and understanding convolutional networks. In Computer Vision–ECCV 2014 (pp. 818-833). Springer International Publishing. 16

  17. Zeiler-Fergus (ZF) “A convnet model that uses the same components (filtering, pooling) but in reverse, so instead of mapping pixels to features does the opposite.” Zeiler, Matthew D., Graham W. Taylor, and Rob Fergus. "Adaptive deconvolutional networks for mid and high level feature learning." Computer Vision (ICCV), 2011 IEEE International Conference on . IEEE, 2011. 17

  18. Zeiler-Fergus (ZF) DeconvN Net Conv et Zeiler, M. D., & Fergus, R. (2014). Visualizing and understanding convolutional networks. In Computer Vision–ECCV 2014 (pp. 818-833). Springer International Publishing. 18

  19. Zeiler-Fergus (ZF) Zeiler, M. D., & Fergus, R. (2014). Visualizing and understanding convolutional networks. In Computer Vision–ECCV 2014 (pp. 818-833). Springer International Publishing. 19

  20. Zeiler-Fergus (ZF) Zeiler, M. D., & Fergus, R. (2014). Visualizing and understanding convolutional networks. In Computer Vision–ECCV 2014 (pp. 818-833). Springer International Publishing. 20

  21. Zeiler-Fergus (ZF): Stride & filter size The smaller stride (2 vs 4) and filter size (7x7 vs 11x11) results in more distinctive features and fewer “dead" features. AlexNet (Layer 1) ZF (Layer 1) 21

  22. Zeiler-Fergus (ZF) Cleaner features in ZF, without the aliasing artifacts caused by the stride 4 used in AlexNet. AlexNet (Layer 2) ZF (Layer 2) 22

  23. Zeiler-Fergus (ZF): Drop out Regularization with more dropout : introduced in the input layer. Hinton, G. E., Srivastava, N., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. R. (2012). Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580 . 23 Chicago

  24. Zeiler-Fergus (ZF): Results 24

  25. Zeiler-Fergus (ZF): Results 25

  26. E2E: Classification: ImageNet ILSRVC ImageNet Classification 2013 -5% Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L. (2015). Imagenet large scale visual recognition challenge. arXiv 26 preprint arXiv:1409.0575 . [web]

  27. E2E: Classification 27

  28. E2E: Classification: GoogLeNet Movie: Inception (2010) 28

  29. E2E: Classification: GoogLeNet ● 22 layers, but 12 times fewer parameters than AlexNet. Szegedy, Christian, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. "Going deeper with convolutions." 29 CVPR 2015. [video] [slides] [poster]

  30. E2E: Classification: GoogLeNet 30

  31. E2E: Classification: GoogLeNet Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014. 31

  32. E2E: Classification: GoogLeNet Multiple scales Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014. 32

  33. E2E: Classification: GoogLeNet (NiN) 3x3 and 5x5 convolutions deal with different scales. Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014. [Slides] 33

  34. E2E: Classification: GoogLeNet Dimensionality reduction Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014. 34

  35. E2E: Classification: GoogLeNet (NiN) 1x1 convolutions does dimensionality reduction (c3<c2) and accounts for rectified linear units (ReLU). Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014. [Slides] 35

  36. E2E: Classification: GoogLeNet In GoogLeNet, the Cascaded 1x1 Convolutions compute reductions before the expensive 3x3 and 5x5 convolutions. 36

  37. E2E: Classification: GoogLeNet Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014. 37

  38. E2E: Classification: GoogLeNet They somewhat spatial invariance, and has proven a benefitial effect by adding an alternative parallel path. 38

  39. E2E: Classification: GoogLeNet Two Softmax Classifiers at intermediate layers combat the vanishing gradient while providing regularization at training time. ...and no fully connected layers needed ! 39

  40. E2E: Classification: GoogLeNet 40

  41. E2E: Classification: GoogLeNet NVIDIA, “NVIDIA and IBM CLoud Support ImageNet Large Scale Visual Recognition Challenge” (2015) 41

  42. E2E: Classification: GoogLeNet Szegedy, Christian, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. "Going deeper with convolutions." CVPR 2015. [video] [slides] [poster] 42

  43. E2E: Classification: VGG Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." International Conference on Learning Representations (2015) . [video] [slides] [project] 43

  44. E2E: Classification: VGG Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." International Conference on Learning Representations (2015) . [video] [slides] [project] 44

  45. E2E: Classification: VGG: 3x3 Stacks Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." International Conference on Learning Representations (2015) . [video] [slides] [project] 45

  46. E2E: Classification: VGG ● No poolings between some convolutional layers. ● Convolution strides of 1 (no skipping). Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." International Conference on Learning Representations (2015) . [video] [slides] [project] 46

  47. E2E: Classification 3.6% top 5 error… with 152 layers !! 47

  48. E2E: Classification: ResNet He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. "Deep Residual Learning for Image Recognition." arXiv preprint arXiv:1512.03385 48 (2015). [slides]

  49. E2E: Classification: ResNet ● Deeper networks (34 is deeper than 18) are more difficult to train. Thin curves: training error Bold curves: validation error He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. "Deep Residual Learning for Image Recognition." arXiv preprint arXiv:1512.03385 (2015). [slides] 49

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Geirhos et al. (2019) Introduction ImageNet classifjcation with CNNs Which image cues are

Geirhos et al. (2019) Introduction ImageNet classifjcation with CNNs Which image cues are

IMAGENET-TRAINED CNNS ARE BIASED TOWARDS TEXTURE; INCREASING SHAPE BIAS IMPROVES ACCURACY AND ROBUSTNESS Geirhos et al. (2019) Introduction ImageNet classifjcation with CNNs Which image cues are learned How infmuential they are

221 views • 18 slides
Modern CNNs Prof. Seungchul Lee Industrial AI Lab. ImageNet Human performance = 5.1 % from

Modern CNNs Prof. Seungchul Lee Industrial AI Lab. ImageNet Human performance = 5.1 % from

Modern CNNs Prof. Seungchul Lee Industrial AI Lab. ImageNet Human performance = 5.1 % from Kaiming He slides &quot;Deep residual learning for image recognition,&quot; ICML, 2016. 2 ImageNet 3 LeNet CNN = Convolutional Neural Networks

792 views • 15 slides
Augmentation Introduction ImageNet Classification with Deep Convolutional Neural Networks,

Augmentation Introduction ImageNet Classification with Deep Convolutional Neural Networks,

Day 2 Lecture 2 Augmentation Introduction ImageNet Classification with Deep Convolutional Neural Networks, Krizhevsky A., 2012 ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) 1.2 million training images , 50,000 validation images, and

537 views • 14 slides
Image as a single label king crab Image Source: ImageNet Image as an object set Man

Image as a single label king crab Image Source: ImageNet Image as an object set Man

Image as a single label king crab Image Source: ImageNet Image as an object set Man Person Woman Woman GIrl Coat King crab Box Image Source: ImageNet Image as a scene graph Man embrace Woman Woman Woman GIrl Relationships:

1.08k views • 90 slides
CNN Architectures ILSVRC: Imagenet Large Scale Visual Recognition Challenge [Russakovsky et al

CNN Architectures ILSVRC: Imagenet Large Scale Visual Recognition Challenge [Russakovsky et al

CS4501: Introduction to Computer Vision CNN Architectures ILSVRC: Imagenet Large Scale Visual Recognition Challenge [Russakovsky et al 2014] The Problem: Classification Classify an image into 1000 possible classes: e.g. Abyssinian cat,

673 views • 36 slides
ImageNet Classification with Deep Convolutional Neural Networks Krizhevsky et. all Outline

ImageNet Classification with Deep Convolutional Neural Networks Krizhevsky et. all Outline

ImageNet Classification with Deep Convolutional Neural Networks Krizhevsky et. all Outline Introduction DataSet Architecture of the Network Reducing overfitting Learning Results Discussion Introduction

1.3k views • 31 slides
ImageNet in 18 minutes for the masses Motivation - training was fast in Google - no technical

ImageNet in 18 minutes for the masses Motivation - training was fast in Google - no technical

ImageNet in 18 minutes for the masses Motivation - training was fast in Google - no technical reason it can't be fast outside of Google - many things are easier procedurally share your jupyter servers, run any code, collaborate with anyone

379 views • 36 slides
Harmonic Analysis of Deep Convolutional Neural Networks Helmut B olcskei Department of

Harmonic Analysis of Deep Convolutional Neural Networks Helmut B olcskei Department of

Harmonic Analysis of Deep Convolutional Neural Networks Helmut B olcskei Department of Information Technology and Electrical Engineering October 2017 joint work with Thomas Wiatowski and Philipp Grohs ImageNet ImageNet ski rock plant

1.36k views • 114 slides
ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky, Ilya Sutskever,

ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky, Ilya Sutskever,

ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton Presented by Tugce Tasci, Kyunghee Kim 05/18/2015 Outline Goal DataSet Architecture of the Network Reducing

709 views • 35 slides
Review on ImageNet Classification with Deep Convolutional Neural Networks by Alex Krizhevsky et. al

Review on ImageNet Classification with Deep Convolutional Neural Networks by Alex Krizhevsky et. al

11/15/2017 Review on ImageNet Classification with Deep Convolutional Neural Networks by Alex Krizhevsky et. al Zhaohui Liang, PhD Candidate, Lassonde School of Engineering Course #: EECS 6412, FALL 2017 Course Name: Data Mining Presenting Date:

299 views • 12 slides
11/21/2018 ImageNet Classification with Deep Convolutional Neural Networks Prepared by Faizaan

11/21/2018 ImageNet Classification with Deep Convolutional Neural Networks Prepared by Faizaan

11/21/2018 ImageNet Classification with Deep Convolutional Neural Networks Prepared by Faizaan Naveed Won Mo (Andy) Jung Lassonde School of Engineering, York University Canada Objective Overview of Convolutional Neural Network

567 views • 17 slides
Training ImageNet in 15 Minutes With ChainerMN: A Scalable Distributed Deep Learning Framework

Training ImageNet in 15 Minutes With ChainerMN: A Scalable Distributed Deep Learning Framework

Training ImageNet in 15 Minutes With ChainerMN: A Scalable Distributed Deep Learning Framework Takuya Akiba, Shuji Suzuki, Keisuke Fukuda, and Kota Uenishi Preferred Networks, Inc. Who are we? Preferred Networks, Inc. (PFN): A Tokyo-based

666 views • 50 slides
ACCELERATED COMPUTING FOR AI Bryan Catanzaro, 28 October 2017 DEEP LEARNING BIG BANG ImageNet

ACCELERATED COMPUTING FOR AI Bryan Catanzaro, 28 October 2017 DEEP LEARNING BIG BANG ImageNet

ACCELERATED COMPUTING FOR AI Bryan Catanzaro, 28 October 2017 DEEP LEARNING BIG BANG ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky Ilya Sutskever Geoffrey E. Hinton University of Toronto University of

269 views • 25 slides
Common Architecture Elements SIGGRAPH Asia Course CreativeAI: Deep Learning for Graphics 1

Common Architecture Elements SIGGRAPH Asia Course CreativeAI: Deep Learning for Graphics 1

Common Architecture Elements SIGGRAPH Asia Course CreativeAI: Deep Learning for Graphics 1 Classification, Segmentation, Detection ImageNet classification performance (for up-to-date top-performers see leaderboards of datasets like ImageNet

707 views • 48 slides
Regionlets for Generic Object Detection A test on ImageNet Tianbao Yang Xiaoyu Wang

Regionlets for Generic Object Detection A test on ImageNet Tianbao Yang Xiaoyu Wang

Regionlets for Generic Object Detection A test on ImageNet Tianbao Yang Xiaoyu Wang Miao Sun University of Missouri Yuanqing Lin Tony X. Han Shenghuo Zhu Introduction Generic object detection is challenging

256 views • 22 slides
Greedy Layerwise Learning Can Scale to ImageNet Edouard Oyallon Eugene Belilovsky, Michael

Greedy Layerwise Learning Can Scale to ImageNet Edouard Oyallon Eugene Belilovsky, Michael

1 Centre for Visual Computing Greedy Layerwise Learning Can Scale to ImageNet Edouard Oyallon Eugene Belilovsky, Michael Eickenberg, Edouard Oyallon &lt;latexit

552 views • 34 slides
Presentation Content 1. Thesis question 2. Current work achieved ETZCR / C.CORNU / 16/06/2010

Presentation Content 1. Thesis question 2. Current work achieved ETZCR / C.CORNU / 16/06/2010

Journe annuelle Clmentine Cornu de prsentation des travaux des 16 th June, 2010 doctorants du ETZCR / C.CORNU / 16/06/2010 LGI2P EUROCOPTER UNCLASSIFIED 1 Presentation Content 1. Thesis question 2. Current work achieved ETZCR /

681 views • 19 slides
1. Key mental capacity issues for COVID-19 management Hospitals discharge to care homes

1. Key mental capacity issues for COVID-19 management Hospitals discharge to care homes

Learning from your experience Taking stock - looking forward National Mental Capacity Forum Fourth Covid-19 Webinar September 9 th 2020 1. Key mental capacity issues for COVID-19 management Hospitals discharge to care homes (e.g.

559 views • 42 slides
Introduction What is data mining? to Data Mining: On what kind of data? Data Mining

Introduction What is data mining? to Data Mining: On what kind of data? Data Mining

Introduction Motivation: Why data mining? Introduction What is data mining? to Data Mining: On what kind of data? Data Mining Data mining functionalities Major issues in data mining 2 Motivation: Necessity is

438 views • 14 slides
Analysis of the Effect of Sample Size on the Quality of Data Mining Models David Watkins SPSS

Analysis of the Effect of Sample Size on the Quality of Data Mining Models David Watkins SPSS

Analysis of the Effect of Sample Size on the Quality of Data Mining Models David Watkins SPSS UK Ltd Overview Question how does sample size affect the quality of a data mining model? Why is this important? Several reasons: Common

354 views • 17 slides
T T o orah Portion Joh John 14:26 Ac Acts ts 9: 9:31 1 C 1 Cori rint nthi hians ans

T T o orah Portion Joh John 14:26 Ac Acts ts 9: 9:31 1 C 1 Cori rint nthi hians ans

T T o orah Portion Joh John 14:26 Ac Acts ts 9: 9:31 1 C 1 Cori rint nthi hians ans 12 12:1-11 11 Ga Galatians 5:22 :22-26 26 The fire shall be burning continually on the altar; it shall not [be allowed to] go out. -

578 views • 9 slides
Disclosures Research Grants: Amgen, AbbVie, Orthotropix, Pfizer, Regeneron, Myosicience

Disclosures Research Grants: Amgen, AbbVie, Orthotropix, Pfizer, Regeneron, Myosicience

6/23/2015 Osteoporosis and Chronic Kidney Disease: Diagnosis and Treatment Recommendations Nancy E. Lane, MD Director, Center for Musculoskeletal Health Endowed Professor of Medicine and Rheumatology University of California at Davis Sacramento,

717 views • 28 slides
Data Monitoring Committee Training Lecture Three: Methods Overview Introduction 1.1 Statistical

Data Monitoring Committee Training Lecture Three: Methods Overview Introduction 1.1 Statistical

Data Monitoring Committee Training Lecture Three: Methods Overview Introduction 1.1 Statistical Methods Overview 1.2 Book References 1.3 DMC Structure and Function Review 1.4 Data Monitoring Rationale 1.5 DMC Recommendations 1.6 Reasons for

452 views • 43 slides
Tiling: A Data Locality Optimizing Algorithm Announcements Monday November 28th, Dr. Sanjay

Tiling: A Data Locality Optimizing Algorithm Announcements Monday November 28th, Dr. Sanjay

Tiling: A Data Locality Optimizing Algorithm Announcements Monday November 28th, Dr. Sanjay Rajopadhye is talking at BMAC Friday December 2nd, Dr. Sanjay Rajopadhye will be leading CS553 Last Monday Kelly &amp; Pugh transformation

398 views • 10 slides

More recommend