Explaining Deep Convolutional Neural Networks via Latent Visual-Semantic Filter Attention

TL;DR

This paper introduces LaViSE, a framework that enables convolutional neural networks to generate textual explanations of their internal filters, aiding interpretability and bias detection without requiring semantic labels during training.

Contribution

LaViSE is a novel modular framework that maps visual features to semantic descriptions at the filter level, applicable to any trained CNN regardless of original training data.

Findings

Generates meaningful descriptions for CNN filters beyond training categories

Enables unsupervised bias analysis in datasets

Works with any trained CNN without needing original training data

Abstract

Interpretability is an important property for visual models as it helps researchers and users understand the internal mechanism of a complex model. However, generating semantic explanations about the learned representation is challenging without direct supervision to produce such explanations. We propose a general framework, Latent Visual Semantic Explainer (LaViSE), to teach any existing convolutional neural network to generate text descriptions about its own latent representations at the filter level. Our method constructs a mapping between the visual and semantic spaces using generic image datasets, using images and category names. It then transfers the mapping to the target domain which does not have semantic labels. The proposed framework employs a modular structure and enables to analyze any trained network whether or not its original training data is available. We show that our method can generate novel descriptions for learned filters beyond the set of categories defined in the training dataset and perform an extensive evaluation on multiple datasets. We also demonstrate a novel application of our method for unsupervised dataset bias analysis which allows us to automatically discover hidden biases in datasets or compare different subsets without using additional labels. The dataset and code are made public to facilitate further research.