Subspace Alignment Based Domain Adaptation for RCNN Detector

TL;DR

This paper introduces a subspace alignment method for domain adaptation of RCNN object detectors, enabling improved detection in new environments without target labels by aligning feature subspaces derived from detections.

Contribution

It presents a novel approach to unsupervised domain adaptation for object detection using subspace alignment based on RCNN detections in unlabeled target data.

Findings

Improved detection accuracy on COCO dataset using the proposed method.

Effective subspace alignment enhances transferability of RCNN detectors.

Method outperforms baseline without target domain labels.

Abstract

In this paper, we propose subspace alignment based domain adaptation of the state of the art RCNN based object detector. The aim is to be able to achieve high quality object detection in novel, real world target scenarios without requiring labels from the target domain. While, unsupervised domain adaptation has been studied in the case of object classification, for object detection it has been relatively unexplored. In subspace based domain adaptation for objects, we need access to source and target subspaces for the bounding box features. The absence of supervision (labels and bounding boxes are absent) makes the task challenging. In this paper, we show that we can still adapt sub- spaces that are localized to the object by obtaining detections from the RCNN detector trained on source and applied on target. Then we form localized subspaces from the detections and show that subspace alignment based adaptation between these subspaces yields improved object detection. This evaluation is done by considering challenging real world datasets of PASCAL VOC as source and validation set of Microsoft COCO dataset as target for various categories.