Alexandros Doumanoglou

Which Direction to Choose? An Analysis on the Representation Power of Self-Supervised ViTs in Downstream Tasks

2025 · Communications in computer and information science

Abstract Self-Supervised Learning (SSL) for Vision Transformers (ViTs) has recently demonstrated considerable potential as a pre-training strategy for a variety of computer vision tasks, including image classification and segmentation, both in standard and few-shot downstream …