"Refining Automated Image Annotations with Tr...

"Refining Automated Image Annotations with Transfer Learning and Dynamic Sampling Strategies"

Author Name : Purma Hriday Reddy, Neerudu Harshitha, Stuti Jajodia, Sahana H J, Sharvesh Subhash, Rishi Satish, Karthikeyan A

ABSTRACT: Data labeling is often a resource-intensive task, demanding significant time and financial investment In many cases, it also necessitates the expertise of domain specialists Active Learning offers a strategy to streamline this process by reducing the number of training data points required, thereby enhancing model performance In this approach, a classification model manages the bulk of data labeling, with human annotators intervening only when needed Active Learning focuses on selecting and annotating the most informative data from a pool of unlabeled data By enabling the system to identify which data points need human labeling, Active Learning can significantly cut down the volume of labeled data necessary for training a model while boosting its accuracy This method prioritizes data points that the model is uncertain about, concentrating on labeling these to expedite the model's learning process