Deep learning models for computer vision often require large amounts of annotated training data, which can be expensive and time-consuming to collect. Data augmentation offers a way to address this challenge. My research focuses on object insertion-based data augmentation techniques that generate synthetic training images by combining objects and backgrounds from existing images. I investigate different strategies for selecting and combining objects, including the use of challenging examples identified through model predictions, to improve model performance and generalization.
Asma Raeisi
PhD student
PhD student specializing in data augmentation for computer vision. My research explores how synthetic training data can improve the performance and generalization of deep learning models.
