ADeLA: Automatic Dense Labeling with Attention for Viewpoint Shift in Semantic Segmentation

We describe a method to deal with performance drop in semantic segmentation caused by viewpoint changes within multi-camera systems, where temporally paired images are readily available, but the annotations may only be abundant for a few typical views. Existing methods alleviate performance drop via...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online) s. 8069 - 8079
Hlavní autoři:	Ren, Hanxiang, Yang, Yanchao, Wang, He, Shen, Bokui, Fan, Qingnan, Zheng, Youyi, Liu, C. Karen, Guibas, Leonidas
Médium:	Konferenční příspěvek
Jazyk:	angličtina
Vydáno:	IEEE 01.06.2022
Témata:	Color grouping and shape analysis; Vision applications and systems Image segmentation Object segmentation Robot vision systems Semantics Shape Transfer/low-shot/long-tail learning; Datasets and evaluation; Image and video synthesis and generation; Machine learning; Robot vision; Scene analysis and understanding; Segmentation Uncertainty
ISSN:	1063-6919
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	We describe a method to deal with performance drop in semantic segmentation caused by viewpoint changes within multi-camera systems, where temporally paired images are readily available, but the annotations may only be abundant for a few typical views. Existing methods alleviate performance drop via domain alignment in a shared space and assume that the mapping from the aligned space to the output is transferable. However, the novel content induced by viewpoint changes may nullify such a space for effective alignments, thus resulting in negative adaptation. Our method works without aligning any statistics of the images between the two domains. Instead, it utilizes a novel attention-based view transformation network trained only on color images to hallucinate the semantic images for the target. Despite the lack of supervision, the view transformation network can still generalize to semantic images thanks to the induced "information transport" bias. Furthermore, to resolve ambiguities in converting the semantic images to semantic labels, we treat the view transformation network as a functional representation of an unknown mapping implied by the color images and propose functional label hallucination to generate pseudo-labels with uncertainties in the target domains. Our method surpasses baselines built on state-of-the-art correspondence estimation and view synthesis methods. Moreover, it outperforms the state-of-the-art unsupervised domain adaptation methods that utilize self-training and adversarial domain alignments. Our code and dataset will be made publicly available.
ISSN:	1063-6919
DOI:	10.1109/CVPR52688.2022.00791