Semantic Validation and Interpretability of Object Detection

Object detection is an aspect of computer vision that can reach high levels of accuracy and is involved in many aspects of public interaction. Accuracy issues can arise when it comes to the application of these models to new data sets; explainability and interpretability of object detection models are still lacking which affects the trust non-experts have in the model's decision-making process. In this paper, we define and test a method of semantic validation for object detection. With this system's implementation, the ability to accurately validate its detections jumps from an original 8% to 73% while still providing interpretability to its users.