The document presents an overview of the ImageNet Challenge, which involves classifying images across 1,000 object categories using a large dataset with 1.2 million training images and 100,000 test images. It discusses advancements in convolutional neural networks (CNNs) such as AlexNet, GoogLeNet, and ResNet, highlighting their architectures and innovations that improved image classification performance. Additionally, it references key research papers and methodologies that contributed to these developments in deep learning for visual recognition.

1. [course site]
Xavier Giro-i-Nieto
xavier.giro@upc.edu
Associate Professor
Universitat Politecnica de Catalunya
Technical University of Catalonia
Image Classification
on ImageNet
#DLUPC

2. 2
ImageNet Challenge
● 1,000 object classes
(categories).
● Images:
○ 1.2 M train
○ 100k test.

3. 3
Russakovsky, Olga, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang et al. "Imagenet
large scale visual recognition challenge." International Journal of Computer Vision 115, no. 3 (2015): 211-252. [web]
ImageNet Dataset

4. Slide credit:
Rob Fergus (NYU)
-9.8%
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., ... & Fei-Fei, L. (2014). Imagenet large scale visual recognition challenge. arXiv
preprint arXiv:1409.0575. [web] 4
Based on SIFT + Fisher Vectors
ImageNet Challenge: 2012

5. AlexNet (Supervision)
5
Orange
A Krizhevsky, I Sutskever, GE Hinton “Imagenet classification with deep convolutional neural networks” NIPS 2012

6. ImageNet Classification 2013
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]
Slide credit:
Rob Fergus (NYU)
6
ImageNet Challenge: 2013

7. The development of better
convnets is reduced to
trial-and-error.
7
Zeiler-Fergus (ZF)
Visualization can help in
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.

8. “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.
8
Zeiler-Fergus (ZF)

9. Zeiler, M. D., & Fergus, R. (2014). Visualizing and understanding convolutional networks. In Computer Vision–ECCV 2014 (pp. 818-833). Springer
International Publishing.
9
Zeiler-Fergus (ZF)

10. 10
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.
Chicago
Zeiler-Fergus (ZF): Drop out

11. ImageNet Classification 2013
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]
-5%
11
ImageNet Challenge: 2013

12. 12NVIDIA, “NVIDIA and IBM CLoud Support ImageNet Large Scale Visual Recognition Challenge” (2015)
ImageNet Challenge: 2014

13. 13
ImageNet Challenge: 2014

14. GoogLeNet (Inception)
14Movie: Inception (2010)

15. 15
22 layers !
Szegedy, Christian, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov,
Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. "Going deeper with convolutions."
GoogLeNet (Inception)

16. 16
GoogLeNet (Inception)

17. 17
Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014.
GoogLeNet (Inception)

18. 18
Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014.
Multiple
scales
GoogLeNet (Inception)

19. GoogLeNet (NiN)
19
3x3 and 5x5 convolutions deal
with different scales.
Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014. [Slides]

20. 20
Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014.
Dimensionality
reduction
GoogLeNet (Inception)

21. 21
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]
GoogLeNet (Inception)

22. 22
In GoogLeNet, the Cascaded 1x1 Convolutions compute reductions before the
expensive 3x3 and 5x5 convolutions.
GoogLeNet (Inception)

23. 23
Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." ICLR 2014.
GoogLeNet (Inception)

24. 24
Two Softmax Classifiers at intermediate layers combat the vanishing gradient while
providing regularization at training time.
...and no fully connected layers needed
(12 times fewer parameters than AlexNet. !)
GoogLeNet (Inception)

25. 25
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]
GoogLeNet (Inception)

26. E2E: Classification: VGG
26
Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." ICLR 2015.
[video] [slides] [project]

27. E2E: Classification: VGG
27
Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition."
International Conference on Learning Representations (2015). [video] [slides] [project]

28. E2E: Classification: VGG: 3x3 Stacks
28
Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image
recognition." International Conference on Learning Representations (2015). [video] [slides] [project]

29. E2E: Classification: VGG
29
Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image
recognition." International Conference on Learning Representations (2015). [video] [slides] [project]
● No poolings between some convolutional layers.
● Convolution strides of 1 (no skipping).

30. 30
3.6% top 5 error…
with 152 layers !!
ImageNet Challenge: 2015

31. E2E: Classification: ResNet
31
He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. "Deep residual learning for image recognition."
CVPR 2016. [slides]

32. E2E: Classification: ResNet
32
● 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."
CVPR 2016. [slides]

33. ResNet
33
● Residual learning: reformulate the layers as learning residual functions with
reference to the layer inputs, instead of learning unreferenced functions
He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. "Deep residual learning for image recognition."
CVPR 2016. [slides]

34. E2E: Classification: ResNet
34
He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. "Deep residual learning for image recognition."
CVPR 2016. [slides]

35. 35
Learn more
Li Fei-Fei, “How we’re teaching computers to understand
pictures” TEDTalks 2014.
Russakovsky, Olga, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang et al. "Imagenet
large scale visual recognition challenge." International Journal of Computer Vision 115, no. 3 (2015): 211-252. [web]

36. 36
The end of the challenge
http://image-net.org/challenges/beyond_ilsvrc

37. 37
Thanks ! Q&A ?
Follow me at
https://imatge.upc.edu/web/people/xavier-giro
@DocXavi
/ProfessorXavi