Showing cs.CVShow all
2 papers · 1 filter
cs.CV2026
Do CNNs Internally Represent Real and Fake Images Differently? A Hidden-Layer Analysis
Moumita Sen Sarma, Pascal Hitzler, Eugene Y. Vasserman
Fake/synthetic images are increasingly prevalent, but it remains unclear whether Convolutional Neural Networks (CNNs) process real and fake images in the same internal manner. This…
cs.CV2026
A Case Study on Concept Induction for Neuron-Level Interpretability in CNN
Moumita Sen Sarma, Samatha Ereshi Akkamahadevi, Pascal Hitzler
Deep Neural Networks (DNNs) have advanced applications in domains such as healthcare, autonomous systems, and scene understanding, yet the internal semantics of their hidden neuron…