Generalizing GradCAM for Embedding Networks
This work addresses the problem of visualizing and explaining predictions for embedding networks, which is important for building trust in AI systems, but it is incremental as it extends an existing method to a new context.
The authors tackled the limitation of existing visualization methods like GradCAM to classification models by introducing EmbeddingCAM, a method that generalizes GradCAM for embedding networks, and demonstrated its effectiveness on the CUB-200-2011 dataset with quantitative and qualitative analysis.
Visualizing CNN is an important part in building trust and explaining model's prediction. Methods like CAM and GradCAM have been really successful in localizing area of the image responsible for the output but are only limited to classification models. In this paper, we present a new method EmbeddingCAM, which generalizes the Grad-CAM for embedding networks. We show that for classification networks, EmbeddingCAM reduces to GradCAM. We show the effectiveness of our method on CUB-200-2011 dataset and also present quantitative and qualitative analysis on the dataset.