About this Event
2461 SW Campus Way, Corvallis, OR 97331
Open Set Learning with Counterfactual Images
In open set recognition, a classifier must label instances of known classes while detecting instances of unknown classes not encountered during training. Counterfactual image generation is a dataset augmentation process based on generative adversarial networks which synthesizes examples that are close to training examples, yet do not belong to any known training category. By augmenting an image classifier with examples generated by this process, the open set recognition problem can be reformulated as classification with an additional class of novel and unknown examples, to improve performance on image classification in the presence of unknown classes.
Co Advisor: Fuxin Li
Co Advisor: Xiaoli Fern
Committee: Weng-Keen Wong
Committee: Alan Fern
GCR: Jimmy Yang
Event Details
See Who Is Interested
0 people are interested in this event
User Activity
No recent activity