Abstract: Decision intelligence (DI) is a methodology entailing data, models, and human expertise combined for assistance in making tough decisions. Its use in image processing over the recent past has gained much traction, opening avenues for improving tasks such as classification, detection, and recognition. The paper compares some of the leading decision intelligence frameworks in image processing. It exposes how some frameworks combine machine learning with deep learning and rule-based decision-making to improve accuracy, efficiency, and interpretability. Comparison is made on benchmark image datasets in different domains for testing based on computational intensity, scalability, robustness, and decision interpretability. It emerged that even though accuracy is higher for computational intensity combined with DI frameworks based on deep learning, hybrid frameworks involving manual input by humans and automatic drawing of inferences are superior in balancing performance with interpretability. The comparison study shows shortcomings and advantages of each framework and where best it can implement in different image processing tasks while also providing support towards decision intelligence research in computer vision.
Indian Journal of Science and Research (Fri,) studied this question.