# Few-Shot Learning with Localization in Realistic Settings

## TL;DR

This paper addresses the challenges of few-shot learning in realistic, heavy-tailed class distributions by introducing new methods that improve localization and feature representation, significantly boosting accuracy in complex real-world scenarios.

## Contribution

The paper presents three novel, parameter-free techniques for few-shot learning in realistic settings, including improved training, object localization, and feature expansion, enhancing model performance.

## Key findings

- Doubled accuracy on the meta-iNat benchmark
- Generalized well to prior benchmarks and complex architectures
- Effective in scenarios with substantial domain shift

## Abstract

Traditional recognition methods typically require large, artificially-balanced training classes, while few-shot learning methods are tested on artificially small ones. In contrast to both extremes, real world recognition problems exhibit heavy-tailed class distributions, with cluttered scenes and a mix of coarse and fine-grained class distinctions. We show that prior methods designed for few-shot learning do not work out of the box in these challenging conditions, based on a new "meta-iNat" benchmark. We introduce three parameter-free improvements: (a) better training procedures based on adapting cross-validation to meta-learning, (b) novel architectures that localize objects using limited bounding box annotations before classification, and (c) simple parameter-free expansions of the feature space based on bilinear pooling. Together, these improvements double the accuracy of state-of-the-art models on meta-iNat while generalizing to prior benchmarks, complex neural architectures, and settings with substantial domain shift.