Thinking Outside the Pool: Active Training Image Creation for Relative Attributes
This work addresses the problem of training data scarcity for fine-grained visual attributes, which is incremental as it builds on active learning by incorporating image generation.
The paper tackles the challenge of curating diverse and informative training images for fine-grained attribute comparisons by proposing an active image generation approach that creates novel realistic images to improve model performance, achieving gains in generalization accuracy on two datasets.
Current wisdom suggests more labeled image data is always better, and obtaining labels is the bottleneck. Yet curating a pool of sufficiently diverse and informative images is itself a challenge. In particular, training image curation is problematic for fine-grained attributes, where the subtle visual differences of interest may be rare within traditional image sources. We propose an active image generation approach to address this issue. The main idea is to jointly learn the attribute ranking task while also learning to generate novel realistic image samples that will benefit that task. We introduce an end-to-end framework that dynamically "imagines" image pairs that would confuse the current model, presents them to human annotators for labeling, then improves the predictive model with the new examples. With results on two datasets, we show that by thinking outside the pool of real images, our approach gains generalization accuracy for challenging fine-grained attribute comparisons.