Taking walks and grocery shopping are certainly the new hobbies of the Corona pandemic. With the majority of our work and personal lives taking place indoors and at our computers, opportunities to get outside are all the more enjoyable. As an enthusiastic walker and environmentalist, I’ve noticed the sudden crowds. Quite naively, I was pleased to see people appreciating nature. But just as Corona has left its mark on our lives, many walkers also leave unwelcome traces in the environment.
TL;DR For the enforcement of Convolutional Neural Networks (CNN), it is crucial that users accept the model and trust in the classification. However, since the networks are not intuitively comprehensible, non-transparent misclassification leads to a loss of trust in the model. To increase comprehensibility, visualization methods can be used that show the relevance of the image regions to non-AI experts. In the following post, some of them are presented and their use is evaluated based on selected evaluation criteria. The main groups of instance-based visualization methods are considered: perturbation-based, activation-based, backpropagation-based and concept-based.
The post was written as part of…
Data Science Master Student @Hamburg