![]() ![]() ![]() Master-CAM utilizes the general localization ability of the Master map to reduce the noise of the maps. In this paper, we propose Master-CAM, which uses Master map to guide multi-scale fusion process to obtain a high-quality class activation map. However, existing methods based on multi-scale fusion cannot effectively remove the noise from larger-scale images. To obtain fine-grained saliency maps, some works fuse saliency signals of the same image at larger scales. The results show that our method obtains a significant improvement compared to the state-of-the-arts.Ĭlass Activation Map (CAM) is one of the most popular approaches to visually explain the convolutional neural networks (CNNs). We conduct extensive experiments to give a comprehensive comparison between the proposed Master-CAM and the state-of-the-art techniques (e.g., Grad-CAM, LayerCAM, and CAMERAS, etc.). Experiments show that Master-Fusion strategy can produce significant performance improvements for other state-of-the-art saliency approaches (e.g., Grad-CAM, and LayerCAM). We present Master-Fusion, a simple yet effective fusion strategy, for fusing different saliency maps for the same input, which is derived from the fusion operation in Master-CAM and can be flexibly applied to other methods. Master-CAM considers the characteristics of saliency maps from multi-scale inputs and uses the map of the original-scale input to guide the multi-scale fusion. ![]() We propose a novel method for high-quality class activation maps, Master-CAM, which utilizes the general localization ability of the Master map to reduce noise from larger-scale inputs.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |