English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Deep neural networks, especially convolutional deep neural networks, arestate-of-the-art methods to classify, segment or even generate images, movies,or sounds. However, these methods lack of a good semantic understanding of whathappens internally. The question, why a COVID-19 detector has classified astack of lung-ct images as positive, is sometimes more interesting than theoverall specificity and sensitivity. Especially when human domain expertknowledge disagrees with the given output. This way, human domain experts couldalso be advised to reconsider their choice, regarding the information pointedout by the system. In addition, the deep learning model can be controlled, anda present dataset bias can be found. Currently, most explainable AI methods inthe computer vision domain are purely used on image classification, where theimages are ordinary images in the visible spectrum. As a result, there is nocomparison on how the methods behave with multimodal image data, as well asmost methods have not been investigated on how they behave when used for objectdetection. This work tries to close the gaps. Firstly, investigating threesaliency map generator methods on how their maps differ across the differentspectra. This is achieved via accurate and systematic training. Secondly, weexamine how they behave when used for object detection. As a practical problem,we chose object detection in the infrared and visual spectrum for autonomousdriving. The dataset used in this work is the Multispectral Object DetectionDataset, where each scene is available in the FIR, MIR and NIR as well asvisual spectrum. The results show that there are differences between theinfrared and visual activation maps. Further, an advanced training with both,the infrared and visual data not only improves the network's output, it alsoleads to more focused spots in the saliency maps.

A Comparison of Deep Saliency Map Generators on Multispectral Data in Object Detection