LSTD: A Low-Shot Transfer Detector for Object Detection

TL;DR

This paper introduces LSTD, a deep learning-based low-shot object detection method that effectively leverages source domain knowledge and novel regularizations to improve detection performance with limited target data.

Contribution

The paper proposes a flexible deep architecture combining SSD and Faster R-CNN, along with a regularized transfer learning framework for low-shot detection.

Findings

LSTD outperforms state-of-the-art methods in low-shot detection tasks.

The architecture effectively integrates SSD and Faster R-CNN advantages.

Regularizations improve fine-tuning with few target images.

Abstract

Recent advances in object detection are mainly driven by deep learning with large-scale detection benchmarks. However, the fully-annotated training set is often limited for a target detection task, which may deteriorate the performance of deep detectors. To address this challenge, we propose a novel low-shot transfer detector (LSTD) in this paper, where we leverage rich source-domain knowledge to construct an effective target-domain detector with very few training examples. The main contributions are described as follows. First, we design a flexible deep architecture of LSTD to alleviate transfer difficulties in low-shot detection. This architecture can integrate the advantages of both SSD and Faster RCNN in a unified deep framework. Second, we introduce a novel regularized transfer learning framework for low-shot detection, where the transfer knowledge (TK) and background depression (BD) regularizations are proposed to leverage object knowledge respectively from source and target domains, in order to further enhance fine-tuning with a few target images. Finally, we examine our LSTD on a number of challenging low-shot detection experiments, where LSTD outperforms other state-of-the-art approaches. The results demonstrate that LSTD is a preferable deep detector for low-shot scenarios.