Frustratingly Simple Few-Shot Object Detection

TL;DR

This paper demonstrates that fine-tuning only the last layer of existing detectors significantly improves few-shot object detection performance, surpassing meta-learning methods and establishing new benchmarks.

Contribution

It introduces a simple fine-tuning approach focusing on the last layer, which outperforms prior meta-learning methods and revises evaluation protocols for more reliable benchmarking.

Findings

Fine-tuning only the last layer is crucial for few-shot detection.

The approach outperforms meta-learning by 2-20 points on benchmarks.

Revised evaluation protocols lead to more stable and reliable comparisons.

Abstract

Detecting rare objects from a few examples is an emerging problem. Prior works show meta-learning is a promising approach. But, fine-tuning techniques have drawn scant attention. We find that fine-tuning only the last layer of existing detectors on rare classes is crucial to the few-shot object detection task. Such a simple approach outperforms the meta-learning methods by roughly 2~20 points on current benchmarks and sometimes even doubles the accuracy of the prior methods. However, the high variance in the few samples often leads to the unreliability of existing benchmarks. We revise the evaluation protocols by sampling multiple groups of training examples to obtain stable comparisons and build new benchmarks based on three datasets: PASCAL VOC, COCO and LVIS. Again, our fine-tuning approach establishes a new state of the art on the revised benchmarks. The code as well as the pretrained models are available at https://github.com/ucbdrive/few-shot-object-detection.