A Framework for Explanation-Aware Visualization and Adjudication in Object Detection: First Results and Perspectives

Context-aware systems require context information, which essentially should rely on high-quality data. Object detection is one particular area enabling context information from the environment to be processed. Ensuring the presence of high-quality data is crucial for machine learning methods to detect objects with high precision. This paper presents a framework for explanation-aware visualization and adjudication in object detection, integrating the user into a semi-automatic verification and adjudication process, where targeted information can be transported by visualization and explanation methods. We discuss a tool for supporting such approaches and present first results and perspectives.