Joint Learning of Generative Translator and Classifier for Visually Similar Classes
This work addresses the challenge of visual classification in data-scarce and visually similar class scenarios, representing an incremental improvement over existing methods.
The paper tackles the problem of improving visual classification accuracy for visually similar classes with scarce data by proposing a Generative Translation Classification Network (GTCN) that jointly trains a classifier and a generative translation network for on-line data augmentation, achieving results where training on 40% of the dataset surpasses baselines trained on the full dataset and full training yields comparable performance to state-of-the-art methods with a lightweight architecture.
In this paper, we propose a Generative Translation Classification Network (GTCN) for improving visual classification accuracy in settings where classes are visually similar and data is scarce. For this purpose, we propose joint learning from a scratch to train a classifier and a generative stochastic translation network end-to-end. The translation network is used to perform on-line data augmentation across classes, whereas previous works have mostly involved domain adaptation. To help the model further benefit from this data-augmentation, we introduce an adaptive fade-in loss and a quadruplet loss. We perform experiments on multiple datasets to demonstrate the proposed method's performance in varied settings. Of particular interest, training on 40% of the dataset is enough for our model to surpass the performance of baselines trained on the full dataset. When our architecture is trained on the full dataset, we achieve comparable performance with state-of-the-art methods despite using a light-weight architecture.