A domain adaptive object detector aims to adapt itself to unseen domains that may contain variations of object appearance, viewpoints or backgrounds. Most existing methods adopt feature alignment either on the image level or instance level. However, image-level alignment on global features may tangle foreground/background pixels at the same time, while instance-level alignment using proposals may suffer from the background noise. Different from existing solutions, we propose a domain adaptation framework that accounts for each pixel via predicting pixel-wise objectness and centerness. Specifically, the proposed method carries out center-aware alignment by paying more attention to foreground pixels, hence achieving better adaptation across domains. We demonstrate our method on numerous adaptation settings with extensive experimental results and show favorable performance against existing state-of-the-art algorithms.
Fig. Proposed framework for domain adaptive object detection. Given the source and target images, we feed them to a shared feature extractor $G$ to obtain their features $F$. Then, the global alignment on these features is performed via a global discriminator $D_{GA}$ and a domain prediction loss $L_{GA}$. Next, we pass the feature through the fully-convolutional module $P$ to produce the classi cation and centerness maps. These maps and the feature $F$ are utilized to generate the center-aware features. Finally, we use a center-aware discriminator $D_{CA}$ and another domain prediction loss $L_{CA}$ to perform the proposed center-aware feature alignment. Note that the bounding box prediction loss $L_{det}$ is only operated on source images using their corresponding ground-truth bounding boxes.
Cheng-Chun Hsu, Yi-Hsuan Tsai, Yen-Yu Lin, and Ming-Hsuan Yang, "Every Pixel Matters: Center-aware Feature Alignment for Domain Adaptive Object Detector", in European Conference on Computer Vision (ECCV), 2020.
@inproceedings{hsu2020epm,
title = {Every Pixel Matters: Center-aware Feature Alignment for Domain Adaptive Object Detector},
author = {Cheng-Chun Hsu, Yi-Hsuan Tsai, Yen-Yu Lin, Ming-Hsuan Yang},
booktitle = {European Conference on Computer Vision},
year = {2020}
}
- • Chen et al. Domain Adaptive Faster R-CNN for Object Detection in the Wild. In CVPR, 2018.
- • Saito et al. Strong-weak distribution alignment for adaptive object detection. In CVPR, 2019.
- • Zhu et al. Adapting object detectors via selective cross-domain alignment. In CVPR, 2019.
- • He et al. Multi-adversarial faster-rcnn for unrestricted object detection. In ICCV, 2019.
- • Kim et al. Diversify and match: A domain adaptive representation learning paradigm for object detection. In CVPR, 2019.
- • Cai et al. Exploring object relation in mean teacher for cross-domain detection. In CVPR, 2019.
- • Hsu et al. Progressive domain adaptation for object detection. In WACV, 2020.
- • Kim et al. Self-training and adversarial background regularization for unsupervised domain adaptive one-stage object detection. In ICCV, 2019.
- • Ren et al. Faster r-cnn: Towards real-time object detection with region proposal networks. In NeurIPS, 2015.
- • Tian et al. Fcos: Fully convolutional one-stage object detection. In ICCV, 2019.