A Case Study on Concept Induction for Neuron-Level Interpretability in CNN

TL;DR

This paper evaluates the generalizability of a concept induction framework for interpreting hidden neurons in CNNs by applying it to the SUN2012 scene recognition dataset, confirming its broader applicability.

Contribution

It demonstrates that a previously proposed neuron interpretability method effectively transfers from ADE20K to SUN2012, showing its potential for wider use in scene understanding.

Findings

Method successfully assigns semantic labels to neurons in SUN2012

Semantic labels are validated through web images and statistical tests

Framework shows broad applicability across datasets

Abstract

Deep Neural Networks (DNNs) have advanced applications in domains such as healthcare, autonomous systems, and scene understanding, yet the internal semantics of their hidden neurons remain poorly understood. Prior work introduced a Concept Induction-based framework for hidden neuron analysis and demonstrated its effectiveness on the ADE20K dataset. In this case study, we investigate whether the approach generalizes by applying it to the SUN2012 dataset, a large-scale scene recognition benchmark. Using the same workflow, we assign interpretable semantic labels to neurons and validate them through web-sourced images and statistical testing. Our findings confirm that the method transfers to SUN2012, showing its broader applicability.