Unsupervised Domain Adaptation from Synthetic to Real Images for Anchorless Object Detection
This work provides an incremental improvement in unsupervised domain adaptation for anchorless object detection, which is beneficial for researchers and practitioners looking to reduce annotation costs by leveraging synthetic data.
This paper addresses the challenge of adapting object detection models trained on synthetic images to perform well on real-world images without real-image annotations. By applying and adapting two unsupervised domain adaptation methods (entropy minimization and maximum squares loss) to an anchorless detector (CenterNet), the authors achieved an increase in mAP from 61% to 69% compared to direct transfer.
Synthetic images are one of the most promising solutions to avoid high costs associated with generating annotated datasets to train supervised convolutional neural networks (CNN). However, to allow networks to generalize knowledge from synthetic to real images, domain adaptation methods are necessary. This paper implements unsupervised domain adaptation (UDA) methods on an anchorless object detector. Given their good performance, anchorless detectors are increasingly attracting attention in the field of object detection. While their results are comparable to the well-established anchor-based methods, anchorless detectors are considerably faster. In our work, we use CenterNet, one of the most recent anchorless architectures, for a domain adaptation problem involving synthetic images. Taking advantage of the architecture of anchorless detectors, we propose to adjust two UDA methods, viz., entropy minimization and maximum squares loss, originally developed for segmentation, to object detection. Our results show that the proposed UDA methods can increase the mAPfrom61 %to69 %with respect to direct transfer on the considered anchorless detector. The code is available: https://github.com/scheckmedia/centernet-uda.