Efficient Solution to Develop an Exercising Algorithm using Model-Centric and Data-Centric Artificially Intelligent Approaches

Data is an integral part of any Artificial Intelligence (AI) based application, and improving the data iteratively can provide more precise outcomes, as opposed to iteratively improving the models. A paradigm shift is underway where the go-to approach for developing AI incorporated products is gradually moving from model-centric to data-centric. The objective of this research work is to provide an efficient solution based on dataset parameters/features by exercising model-centric and/or data-centric AI approaches. In this novel method, data is subjected to both the approaches, either individually or to their hybridizations, and the results are compared quantitatively against a common ground. The CIFAR-10 dataset has been processed by the proposed method and concluded that the model-centric approach gives better performance than the data-centric approach with the same hyperparameters. The solution proposed in this work can be employed to develop an automated framework for image classification. Similarly, such frameworks can be developed for other computer vision-based tasks such as object detection, object localization, and so on.