Hallucination Improves Few-Shot Object Detection
This addresses the challenge of detecting novel objects with very few annotated examples, which is important for practical applications, though it is an incremental improvement on existing methods.
The paper tackles the problem of few-shot object detection with extremely limited examples by introducing a hallucinator network that generates additional training examples in feature space, achieving new state-of-the-art performance on the COCO benchmark.
Learning to detect novel objects from few annotated examples is of great practical importance. A particularly challenging yet common regime occurs when there are extremely limited examples (less than three). One critical factor in improving few-shot detection is to address the lack of variation in training data. We propose to build a better model of variation for novel classes by transferring the shared within-class variation from base classes. To this end, we introduce a hallucinator network that learns to generate additional, useful training examples in the region of interest (RoI) feature space, and incorporate it into a modern object detection model. Our approach yields significant performance improvements on two state-of-the-art few-shot detectors with different proposal generation procedures. In particular, we achieve new state of the art in the extremely-few-shot regime on the challenging COCO benchmark.