Fast Hierarchical Learning for Few-Shot Object Detection

TL;DR

This paper introduces a hierarchical learning approach for few-shot object detection that reduces training time and prevents catastrophic forgetting, achieving competitive results on MS-COCO and enabling class-refined detection.

Contribution

It proposes a novel hierarchical learning framework with a specialized training strategy that improves efficiency and preserves base class performance in few-shot detection.

Findings

Achieves competitive performance on MS-COCO benchmark.

Significantly reduces training time compared to SGD.

Maintains base class accuracy while learning new classes.

Abstract

Transfer learning based approaches have recently achieved promising results on the few-shot detection task. These approaches however suffer from ``catastrophic forgetting'' issue due to finetuning of base detector, leading to sub-optimal performance on the base classes. Furthermore, the slow convergence rate of stochastic gradient descent (SGD) results in high latency and consequently restricts real-time applications. We tackle the aforementioned issues in this work. We pose few-shot detection as a hierarchical learning problem, where the novel classes are treated as the child classes of existing base classes and the background class. The detection heads for the novel classes are then trained using a specialized optimization strategy, leading to significantly lower training times compared to SGD. Our approach obtains competitive novel class performance on few-shot MS-COCO benchmark, while completely retaining the performance of the initial model on the base classes. We further demonstrate the application of our approach to a new class-refined few-shot detection task.