Extending and Analyzing Self-Supervised Learning Across Domains
This work addresses the gap in self-supervised learning for non-standard domains, providing insights for researchers in fields like satellite or biological imagery, though it is incremental as it extends existing methods to new data.
The paper tackled the problem of applying self-supervised learning methods to diverse, smaller domains beyond ImageNet, finding that Rotation is the most semantically meaningful task while other methods rely more on distributional properties, and identifying areas like fine-grain classification where all tasks underperform.
Self-supervised representation learning has achieved impressive results in recent years, with experiments primarily coming on ImageNet or other similarly large internet imagery datasets. There has been little to no work with these methods on other smaller domains, such as satellite, textural, or biological imagery. We experiment with several popular methods on an unprecedented variety of domains. We discover, among other findings, that Rotation is by far the most semantically meaningful task, with much of the performance of Jigsaw and Instance Discrimination being attributable to the nature of their induced distribution rather than semantic understanding. Additionally, there are several areas, such as fine-grain classification, where all tasks underperform. We quantitatively and qualitatively diagnose the reasons for these failures and successes via novel experiments studying pretext generalization, random labelings, and implicit dimensionality. Code and models are available at https://github.com/BramSW/Extending_SSRL_Across_Domains/.