Low-Shot Learning from Imaginary Data

Abstract

Humans can quickly learn new visual concepts, perhaps because they can easily visualize or imagine what novel objects look like from different views. Incorporating this ability to hallucinate novel instances of new concepts might help machine vision systems perform better low-shot learning, i.e., learning concepts from few examples. We present a novel approach to low-shot learning that uses this idea. Our approach builds on recent progress in meta-learning ("learning to learn") by combining a meta-learner with a "hallucinator" that produces additional training examples, and optimizing both models jointly. Our hallucinator can be incorporated into a variety of meta-learners and provides significant gains: up to a 6 point boost in classification accuracy when only a single training example is available, yielding state-of-the-art performance on the challenging ImageNet low-shot classification benchmark.

CVPR 2018 camera-ready version

References in corpus (3)

• Prototypical Networks for Few-shot Learning
• A Simple Way to Initialize Recurrent Networks of Rectified Linear Units
• Good Semi-supervised Learning that Requires a Bad GAN

Cited by in corpus (35)

• Learning to Self-Train for Semi-Supervised Few-Shot Classification
• Free Lunch for Few-shot Learning: Distribution Calibration
• Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need?
• Dynamic Few-Shot Visual Learning without Forgetting
• Large-Scale Long-Tailed Recognition in an Open World
• Prior Guided Feature Enrichment Network for Few-Shot Segmentation
• Rectifying the Shortcut Learning of Background for Few-Shot Learning
• Few-Shot Object Detection with Attention-RPN and Multi-Relation Detector
• Edge-labeling Graph Neural Network for Few-shot Learning
• Low-shot Learning via Covariance-Preserving Adversarial Augmentation Networks
• Restoring Negative Information in Few-Shot Object Detection
• Adversarial Feature Hallucination Networks for Few-Shot Learning
• Few-Shot Learning via Saliency-guided Hallucination of Samples
• Synthesizing lesions using contextual GANs improves breast cancer classification on mammograms
• A Broader Study of Cross-Domain Few-Shot Learning
• Conditional Infilling GANs for Data Augmentation in Mammogram Classification
• Generative One-Shot Face Recognition
• Learning from the Past: Continual Meta-Learning via Bayesian Graph Modeling
• PFENet++: Boosting Few-shot Semantic Segmentation with the Noise-filtered Context-aware Prior Mask
• Dizygotic Conditional Variational AutoEncoder for Multi-Modal and Partial Modality Absent Few-Shot Learning
• Machine learning with limited data
• Image-Level or Object-Level? A Tale of Two Resampling Strategies for Long-Tailed Detection
• Adaptive Cross-Modal Few-Shot Learning
• How good is my GAN?
• Fine-grained Image-to-Image Transformation towards Visual Recognition
• Towards Robust Pattern Recognition: A Review
• Deep Low-Shot Learning for Biological Image Classification and Visualization from Limited Training Samples
• A Strong Baseline for Semi-Supervised Incremental Few-Shot Learning
• Learning Classifiers for Domain Adaptation, Zero and Few-Shot Recognition Based on Learning Latent Semantic Parts
• Generative Modeling for Multi-task Visual Learning
• Learning to Learn Image Classifiers with Visual Analogy
• One-Shot Image Classification by Learning to Restore Prototypes
• One-Shot Induction of Generalized Logical Concepts via Human Guidance
• Depth Guided Adaptive Meta-Fusion Network for Few-shot Video Recognition
• Regularising Deep Networks with Deep Generative Models