Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
Paper summary This paper attempts to understand the representations learnt by deep convolutional neural networks by introducing two interpretable visualization techniques. Main contributions: - Class model visualizations - These are obtained by making numerical optimizations in the input space to maximize the class score. Gradients are calculated wrt input and are used to update the input image (initialized with zero image), while weights are kept fixed to those obtained from training. - Image-specific saliency map visualizations - These are approximated by using the same gradient as before (gradient of class score wrt input). The absolute pixel-wise max across channels produces the saliency map. - Relation between DeconvNet and optimization-based visualizations - Visualizations using DeconvNet are the same as gradient-based methods except for ReLU. In regular backprop, gradients flow through ReLU to units with positive input activations, whereas in case of a DeconvNet, it is computed on positive output reconstructions. ## Strengths - The visualization techniques are simple ideas and the results are interpretable. They show that the method proposed by Erhan et al. in an unsupervised setting is useful to CNNs trained in a supervised manner as well. - The image-specific class saliency can be interpreted as those pixels which need to be changed the least to have a maximum impact on the classification score. - The relation between DeconvNet visualizations and optimization-based visualizations is insightful. ## Weaknesses / Notes - The thinking behind initializing with zero image and L2 regularization in class model visualizations was missing.
arxiv.org
scholar.google.com
Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
Simonyan, Karen and Vedaldi, Andrea and Zisserman, Andrew
arXiv e-Print archive - 2013 via Local Bibsonomy
Keywords: dblp


Summary from Shagun Sodhani
Loading...
Your comment:
Summary from openreview
Loading...
Your comment:
Summary from abhshkdz
Loading...
Your comment:


ShortScience.org allows researchers to publish paper summaries that are voted on and ranked!
About

Sponsored by: and