Challenges in Disentangling Independent Factors of Variation
This work addresses the challenge of disentangling factors in models for tasks like image synthesis and classification, but it is incremental as it builds on existing autoencoder methods and focuses on specific weak labeling scenarios.
The paper tackles the problem of building models that disentangle independent factors of variation using weakly labeled data, where labels indicate what factor changed between samples but not the magnitude, and demonstrates that the proposed autoencoder model can successfully transfer attributes on several datasets while highlighting cases where reference ambiguity occurs.
We study the problem of building models that disentangle independent factors of variation. Such models could be used to encode features that can efficiently be used for classification and to transfer attributes between different images in image synthesis. As data we use a weakly labeled training set. Our weak labels indicate what single factor has changed between two data samples, although the relative value of the change is unknown. This labeling is of particular interest as it may be readily available without annotation costs. To make use of weak labels we introduce an autoencoder model and train it through constraints on image pairs and triplets. We formally prove that without additional knowledge there is no guarantee that two images with the same factor of variation will be mapped to the same feature. We call this issue the reference ambiguity. Moreover, we show the role of the feature dimensionality and adversarial training. We demonstrate experimentally that the proposed model can successfully transfer attributes on several datasets, but show also cases when the reference ambiguity occurs.