Papperoni

2014

Visualizing and Understanding Convolutional Networks

Matthew D. Zeiler, Rob Fergus

Open PDF Google Scholar

citations

Cite Score

93

AI summary

The paper introduces a novel visualization technique using Deconvolutional Networks (deconvnet) to understand the internal operation of Convolutional Networks (convnet). It diagnoses potential problems with the model and explores the model's generalization ability on the ImageNet, Caltech-101, and Caltech-256 datasets.

Main Contributions

Introduces a novel visualization technique using a multi-layered Deconvolutional Network (deconvnet) to map feature activations back to the input pixel space.
Performs sensitivity analysis of the classifier output by occluding portions of the input image.
Finds model architectures that outperform (Krizhevsky et al., 2012) on the ImageNet classification benchmark.
Shows ImageNet model generalizes well to other datasets such as Caltech-101 and Caltech-256.
Achieves state-of-the-art results on Caltech-101 and Caltech-256 datasets.

Abstract

Large Convolutional Network models have recently demonstrated impressive classification performance on the ImageNet benchmark (Krizhevsky et al., 2012). However there is no clear understanding of why they perform so well, or how they might be improved. In this paper we address both issues. We introduce a novel visualization technique that gives insight into the function of intermediate feature layers and the operation of the classifier. Used in a diagnostic role, these visualizations allow us to find model architectures that outperform Krizhevsky et al. on the ImageNet classification benchmark. We also perform an ablation study to discover the performance contribution from different model layers. We show our ImageNet model generalizes well to other datasets: when the softmax classifier is retrained, it convincingly beats the current state-of-the-art results on Caltech-101 and Caltech-256 datasets.

Citation Graph

Loading graph...

References [23]

Sort:

Filter:

[1]ImageNet Classification With Deep Convolutional Neural Networks

Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton - 2012

71 papers in library cite

I'm giving this a 5 just because of the impact, but this is VEEERY derivative of earlier work. Kudos for them for putting it all together, but really there's nothing revolutionary here.

[2]A Fast Learning Algorithm for Deep Belief Nets

Geoffrey E. Hinton, S. Osindero, Y. Teh - 2006

43 papers in library cite

The paper does not explain anything. It just throws the idea and a bunch of math, but doesn't really care to explain the concepts.

[3]Backpropagation Applied to Handwritten Zip-Code Recognition

Yann Lecun, B. Boser, John S. Denker, D. Henderson, R. E. Howard, W. Hubbard, L. D. Jackal - 1989

24 papers in library cite

The first convolution NN! Very simple concept and very simply explained. Very good results and overall a good read.

[4]Improving Neural Networks by Preventing Co-Adaptation of Feature Detectors

Geoffrey E. Hinton, N. Srivastava, Alex Krizhevsky, Ilya Sutskever, Ruslan R. Salakhutdinov - 2012

25 papers in library cite

Dropout, super impactful. The idea that you are training many estimators at once is also very nice.

[5]Extracting and Composing Robust Features With Denoising Autoencoders

Pascal Vincent, Hugo Larochelle, Yoshua Bengio, Pierre Antoine Manzagol - 2008

25 papers in library cite

I am *so* glad we found an alternative to DBNs. Also, introduced the idea of denoising which is nice.

[6]Greedy Layer-Wise Training of Deep Networks

Yoshua Bengio, P. Lamblin, D. Popovici, Hugo Larochelle - 2006

33 papers in library cite

Bengio is perfect. This is everything that Hinton's paper hoped to be. Very well explained, and also tying back to real use cases (not just "hey, the math works and it reduced the score")

[7]Multi-Column Deep Neural Networks for Image Classification

Dan C. Ciresan, Ueli Meier, Jürgen Schmidhuber - 2012

11 papers in library cite

Very nice paper! And I am impressed they used CNNs before Hinton's paper. It's a shame there are so few citations. They also propose max-pooling and actually give a good explanation about it.

[8]DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition

J. Donahue, Y. Jia, Oriol Vinyals, J. Hoffman, N. Zhang, E. Tzeng, Trevor Darrell - 2014

15 papers in library cite

Very nice paper. First I've seen (and based on the text, first ever) about feature extraction for images. It's very nice to see embeddings doing SotA

[9]One-Shot Learning of Object Categories

Li Fei Fei, Rob Fergus, Pietro Perona - 2006

5 papers in library cite

It's a nice problem they are trying to solve, though I doubt the effectiveness. Maybe it was really relevant in a time where they didn't have data. Problem is it focuses too much on the methodology, and I am not too interested in it.

[10]What Is the Best Multi-Stage Architecture for Object Recognition?

K. Jarrett, Koray Kavukcuoglu, Marc'aurelio Ranzato, Yann Lecun - 2009

20 papers in library cite

So boring and convoluted... They try to make things more abstract by saying "filter banks" and "2 stages", but in the end it's CNNs. It seems like they are trying to tie NNs to the rest of ML, but it doesn't work.

[11]Visualizing Higher-Layer Features of a Deep Network

Dumitru Erhan, Yoshua Bengio, Aaron Courville, Pascal Vincent - 2009

4 papers in library cite

Very nice the way that they tackled it as an optimization problem using gradient descent. I think this is a similar approach to adversarial examples (not sure if this is what inspired them, I don't remember)

[12]Adaptive Deconvolutional Networks for Mid and High Level Feature Learning

M. Zeiler, G. Taylor, Rob Fergus - 2011

2 papers in library cite

The idea is very nice but I think this is soooo poorly explained. I had to ask AI to explain it to me - isn't that crazy? It isn't even a hard concept!

[13]Histograms of Oriented Gradients for Human Detection

N. Dalal, B. Triggs - 2005

12 papers in library cite

[14]Linear Spatial Pyramid Matching Using Sparse Coding for Image Classification

Jihan Yang, K. Yu, Y. Gong, T. Huang - 2009

8 papers in library cite

[15]Unbiased Look at dataset Bias

Antonio Torralba, A. Efros - 2011

5 papers in library cite

[16]Tiled Convolutional Neural Networks

Quoc V. Le, J. Ngiam, Ziru Chen, D. Chia, P. W. Koh, Andrew Y. Ng - 2010

4 papers in library cite

[17]The Caltech 256

Greg Griffin, Alex Holub, Pietro Perona - 2006

3 papers in library cite

[18]Classification Entry

N. Gunji, T. Higuchi, K. Yasumoto, H. Muraoka, Y. Ushiku, T. Harada, Y. Kuniyoshi - 2012

1 paper in library cites

[19]Efficient Learning of Sparse, Distributed, Convolutional Feature Representations for Object Recognition

K. Sohn, D. Jung, Honglak Lee, A. I. Hero - 2011

1 paper in library cites

[20]Generalized Hierarchical Matching for Sub-Category Aware Object Classification

Shuicheng Yan, J. Dong, Qinlang Chen, Z. Song, Y. Pan, W. Xia, Zhongqiang Huang, Y. Hua, S. Shen - 2012

1 paper in library cites

[21]Hybrid Coding for Selective Search

K. Sande, J. Uijlings, C. Snoek, A. Smeulders - 2012

1 paper in library cites

[22]Multipath Sparse Coding Using Hierarchical Matching Pursuit

L. Bo, Xiang Ren, D. Fox - 2013

1 paper in library cites

[23]On the Analysis and Interpretation of Inhomogeneous Quadratic Forms as Receptive Fields

P. Berkes, L. Wiskott - 2006

1 paper in library cites

Cited by

15

papers in your library

Cites

12

papers in your library

Read

on July 31, 2025

Very good explanation and visualization of CNNs, and also nice that they use their findings to improve the performance. The ablation study is also nice.

Tags

Paper Aliases

No aliases