Dear all,
it is my great pleasure to forward a master thesis talk invitation given by Talha Munir from Fraunhofer EAS in Dresden. You can find title, abstract and MS Teams link below.
Best regards
Johannes
Explainable AI: Improving the clarity of attention maps by differentiable binarization
Time: 29.09., 15:00
Abstract:
Visual explanation enables humans to understand the decision making process of neural networks. The attention branch network architecture is a promising approach for explainable neural networks. It generates attention maps highlighting the important regions in the image for classification. In my thesis, investigations are being carried out to what extent the significance of attention map values can be further increased. As an approach, we have applied multiple loss functions over attention values and visually evaluated the resulting attention maps. I will present this approach and preliminary results in my Master’s Seminar talk and I am looking forward to your suggestions on how to further enhance the clarity in the attention maps for explainability.
Click HERE to attend.