Coarse Grained Label

Coarse-grained labels, representing broader categories rather than fine-grained distinctions, are increasingly used in machine learning due to their ease of acquisition. Current research focuses on leveraging these coarse labels to improve performance on fine-grained tasks, employing techniques like multiple instance learning (MIL), and developing novel architectures that effectively integrate coarse and fine-grained information, often incorporating self-supervised learning and attention mechanisms. This work is significant because it addresses the limitations of fully supervised learning in data-scarce scenarios, particularly in fields like medical image analysis and remote sensing, where obtaining detailed annotations is costly and time-consuming.