Hallucination Improves Few-Shot Object Detection
Authors:
- 1 June 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 13003-13012
Abstract
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.Keywords
All Related Versions
This publication has 24 references indexed in Scilit:
- Learning to Learn: Model Regression Networks for Easy Small Sample Learning2016
- You Only Look Once: Unified, Real-Time Object Detection2016
- Fast R-CNN2015
- Model recommendation: Generating object detectors from few samples2015
- ImageNet Large Scale Visual Recognition ChallengeInternational Journal of Computer Vision, 2015
- Spatial Pyramid Pooling in Deep Convolutional Networks for Visual RecognitionIEEE Transactions on Pattern Analysis and Machine Intelligence, 2015
- Capturing Long-Tail Distributions of Object Subcategories2014
- Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation2014
- The Pascal Visual Object Classes (VOC) ChallengeInternational Journal of Computer Vision, 2009
- One-shot learning of object categoriesIEEE Transactions on Pattern Analysis and Machine Intelligence, 2006