Mixture of Teacher Experts for Source-Free Domain Adaptive Object Detection

Unsupervised domain adaptive object detection methods transfer knowledge from the labelled source domain to a visually distinct and unlabeled target domain. Most methods achieve this by training the detector model with the help of both labeled source and unlabeled target data. However, in real-world scenarios, gaining access to source data is not practical due to privacy concerns, legal issues and inefficient data transmission. To this end, we tackle the problem of Source-Free Domain Adaptive Object Detection, where during adaptation, we do not have access to the source data but only the source trained model. Specifically, we introduce Mixture of Teacher Experts (MoTE) method, where our key idea is to exploit the prediction uncertainty through a mixture of teacher models and progressively train the student model. We evaluate the proposed method by conducting extensive experiments on several object detection benchmark datasets to demonstrate the effectiveness of the proposed mixture of teacher expert based student-teacher training, specifically for source-free adaptation.

1. INTRODUCTION

In the past decade, Convolutional Neural Network (CNN) based methods have achieved significant improvements for the task of object detection [1], [2], [3], [4], [5], [6]. These advancements have led to several successful applications in the real-world like video surveillance, autonomous navigation, image analysis, etc. However, existing object detectors are highly data-driven and susceptible to data distribution shift/domain shift. Consider a real-world application such as an autonomous car, where the object detectors are likely trained on data from one particular city, e.g. Tokyo City (TC), also referred to as source domain. While deploying the same model in a different city, e.g., New York City (NYC), also referred to as the target domain, the detector experiences severe performance degradation. This is due to the fact that object appearance, scene type, illumination, background, or weather condition in NYC are visually distinct from the TC. This problem is widely studied under the Unsupervised Domain Adaption (UDA) setting [7, 8, 9, 10, 11, 12, 13, 14], where the detection model is adopted to transfer the knowledge from the labelled source (e.g. TC) to unlabeled target (e.g. NYC) domain to overcome this poor generalization problem.