Systematic Review on Transfer Learning in Image Recognition and Their Applications

Abstract

Deep learning is a fast-growing field of study that addresses difficult issues with increasingly intricate neural network topologies. A more modern branch of deep learning called transfer learning emphasizes the usefulness of knowledge transfer from one machine learning task to another, particularly in image classification.Research on transfer learning in image recognition is currently very active and dynamic. Utilizing pre-trained models, frequently based on convolutional neural networks (CNNs), and applying them to additional CNNs is a common application of transfer learning. Compared to starting from scratch, this method has produced notable improvements in accuracy, efficiency, and model training facilitation. All things considered, the review paper offers insightful information about the status of transfer learning in image recognition today, its significance, and potential directions for future study.