5 papers
Tied Prototype Model for Few-Shot Medical Image Segmentation
Hyeongji Kim, Stine Hansen, Michael Kampffmeyer
Common prototype-based medical image few-shot segmentation (FSS) methods model foreground and background classes using class-specific prototypes. However, given the high variabilit…
Inspecting class hierarchies in classification-based metric learning models
Hyeongji Kim, Pekka Parviainen, Terje Berge +1
Most classification models treat all misclassifications equally. However, different classes may be related, and these hierarchical relationships must be considered in some classifi…
Distance-Ratio-Based Formulation for Metric Learning
Hyeongji Kim, Pekka Parviainen, Ketil Malde
In metric learning, the goal is to learn an embedding so that data points with the same class are close to each other and data points with different classes are far apart. We propo…
Measuring Adversarial Robustness using a Voronoi-Epsilon Adversary
Hyeongji Kim, Pekka Parviainen, Ketil Malde
Previous studies on robustness have argued that there is a tradeoff between accuracy and adversarial accuracy. The tradeoff can be inevitable even when we neglect generalization. W…
Beyond image classification: zooplankton identification with deep vector space embeddings
Ketil Malde, Hyeongji Kim
Zooplankton images, like many other real world data types, have intrinsic properties that make the design of effective classification systems difficult. For instance, the number of…