Explaining YOLO: Leveraging Grad-CAM to Explain Object Detections
This work addresses the need for explainability in object detection systems, but it is incremental as it applies an existing method (Grad-CAM) to a new model (YOLO).
The paper tackled the problem of explainability for visual object detectors by integrating Grad-CAM into the YOLO architecture to compute attribution-based explanations for individual detections, finding that normalization significantly impacts interpretation.
We investigate the problem of explainability for visual object detectors. Specifically, we demonstrate on the example of the YOLO object detector how to integrate Grad-CAM into the model architecture and analyze the results. We show how to compute attribution-based explanations for individual detections and find that the normalization of the results has a great impact on their interpretation.