Paper accepted at international research conference (IEEE ISBI 2024, Athens), March 13, 2024
Our research paper "Integrating kNN with Foundation Models for Adaptable and Privacy-Aware Image Classification" was evaluated very positively by independent experts and accepted for presentation at the IEEE ISBI 2024 in Athens. The International Symposium for Biomedical Imaging (ISBI) is a leading annual international research conference in the field of medical imaging.
We are delighted about this great team achievement by Sebastian D?rrich, Tobias Archut and Francesco Di Salvo. It should be particularly emphasized that Tobias Archut made a significant contribution to this work with his outstanding bachelor thesis (supervisor: Sebastian D?rrich).
Overview:
Traditional deep learning models encode knowledge within their parameters, limiting transparency and adaptability to data changes. This poses challenges for addressing user data privacy concerns. To overcome this limitation, we propose to store embeddings of the training data independently of the model weights. Our approach integrates the k-Nearest Neighbor (k-NN) classifier with a vision-based foundation model pre-trained on natural images in a self-supervised manner, enhancing interpretability and adaptability while addressing privacy concerns.
Key Features:
- Integration of k-NN classifier with recent vision-based foundation models
- Flexible data storage system for dynamic data modifications without retraining
- Thorough assessment across established benchmarks and tasks including continual learning and data removal scenarios
