Feature CAM: Interpretable AI in Image Classification

arXiv:2403.05658v1 Announce Type: new
Abstract: Deep Neural Networks have often been called the black box because of the complex, deep architecture and non-transparency presented by the inner layers. There is a lack of trust to use Artificial Intelligence in critical and high-precision fields such as security, finance, health, and manufacturing industries. A lot of focused work has been done to provide interpretable models, intending to deliver meaningful insights into the thoughts and behavior of neural networks. In our research, we …

abstract architecture artificial artificial intelligence arxiv black box box classification cs.ai cs.cv cs.mm feature fields finance health image industries intelligence interpretable ai manufacturing networks neural networks precision security transparency trust type