Delta-encoder: an effective sample synthesis method for few-shot object recognition

TL;DR

This paper introduces Delta-encoder, a modified auto-encoder that synthesizes new samples for unseen categories in few-shot object recognition, significantly improving classification performance with minimal examples.

Contribution

The paper presents Delta-encoder, a novel auto-encoder-based method that learns to generate synthetic samples for unseen classes using intra-class deformations, enhancing few-shot recognition.

Findings

Outperforms state-of-the-art in one-shot recognition

Effective in few-shot scenarios with limited data

Synthesizes realistic samples for unseen categories

Abstract

Learning to classify new categories based on just one or a few examples is a long-standing challenge in modern computer vision. In this work, we proposes a simple yet effective method for few-shot (and one-shot) object recognition. Our approach is based on a modified auto-encoder, denoted Delta-encoder, that learns to synthesize new samples for an unseen category just by seeing few examples from it. The synthesized samples are then used to train a classifier. The proposed approach learns to both extract transferable intra-class deformations, or "deltas", between same-class pairs of training examples, and to apply those deltas to the few provided examples of a novel class (unseen during training) in order to efficiently synthesize samples from that new class. The proposed method improves over the state-of-the-art in one-shot object-recognition and compares favorably in the few-shot case. Upon acceptance code will be made available.