Research

Based on mathematical and statistical modeling of visual representations and on large scale experimental work our collaboratory strives to make foundational contributions to basic research areas such as:

Explainability – Assist human decision making with AI in fine-grained medical domains. Can we construct AI systems for fine-grained classification that are causally explainable and interpretable by humans?

Self-supervised learning – Construct models that learn representations rich enough for fine-grained classification without supervision. Based on expert knowledge we will exploit self-supervised learning techniques to establish representations for fine-grained few shot generalization.

Novelty detection – Develop scalable novelty detection that is robust against extreme class imbalance and other effects of long-tailed distributions in real-world fine-grained classification problems.

Fine-Grained Analysis

Monarch Butterfly

Monarch Butterfly [1]

Viceroy Butterfly

Viceroy Butterfly [2]

Fine-grained categorization refers to the challenge of distinguishing between closely related entities that may appear very similar at first glance. For example, consider the task of identifying distinguishing features between the Monarch butterfly and the Viceroy butterfly. Despite their visual similarities, subtle differences, such as wing patterns and color variations, provide important semantic information. The goal of fine-grained analysis is to extract and leverage these nuanced details, even when they are subtle or easily overlooked.

This concept of fine-grained analysis is applicable across a diverse range of domains. In medical data, for instance, distinguishing between slightly different types of lesions in medical imaging can significantly impact diagnosis and treatment. In the realm of language sentiment, understanding the nuanced emotional tone in text—such as distinguishing between sarcasm and genuine praise—can improve natural language processing systems. Similarly, in remote sensing, differentiating between types of vegetation or detecting minor changes in environmental conditions relies on fine-grained analysis.

Achieving advancements in fine-grained analysis requires not only cutting-edge machine learning research but also a deep integration with, and respect for, human expertise and domain-specific knowledge. Collaboration with experts from various fields ensures that the subtle distinctions we aim to identify are meaningful and relevant. This multidisciplinary approach helps bridge the gap between advanced computational techniques and practical, real-world applications.

Below we showcase some of the applications we are interrested in.

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.

References

James St. John, CC BY 2.0, via Wikimedia Commons
Thomas Shahan 3, CC BY 2.0, via Wikimedia Commons

Research

Fine-Grained Analysis

Some application related to brains

Some application related to EHR

Some application related to Plants (Wheat?)

References