Scaling Semantic Segmentation Beyond 1K Classes on a Single GPU

Jain, ShipraKTH,Skolan för elektroteknik och datavetenskap (EECS),Swiss Fed Inst Technol, Comp Vis Lab, Zurich, Switzerland. (author)

Part of proceedings: ISBN 978-1-6654-2812-5QC 20220715
The state-of-the-art object detection and image classification methods can perform impressively on more than 9k classes. In contrast, the number of classes in semantic segmentation datasets is relatively limited. This is not surprising when the restrictions caused by the lack of labeled data and high computation demand for segmentation are considered. In this paper, we propose a novel training methodology to train and scale the existing semantic segmentation models for a large number of semantic classes without increasing the memory overhead. In our embedding-based scalable segmentation approach, we reduce the space complexity of the segmentation model's output from O

Paudel, Danda PaniSwiss Fed Inst Technol, Comp Vis Lab, Zurich, Switzerland. (author)
Danelljan, MartinSwiss Fed Inst Technol, Comp Vis Lab, Zurich, Switzerland. (author)
Van Gool, LucSwiss Fed Inst Technol, Comp Vis Lab, Zurich, Switzerland.;Katholieke Univ Leuven, Leuven, Belgium. (author)
KTHSkolan för elektroteknik och datavetenskap (EECS) (creator_code:org_t)

In:Proceedings of the IEEE International Conference on Computer Vision: Institute of Electrical and Electronics Engineers (IEEE), s. 7406-7416

By the author/editor: Jain, Shipra; Paudel, Danda Pa ...; Danelljan, Marti ...; Van Gool, Luc

About the subject

NATURAL SCIENCES: NATURAL SCIENCES; and Computer and Inf ...; and Computer Science ...

Kungliga biblioteket hanterar dina personuppgifter i enlighet med EU:s dataskyddsförordning (2018), GDPR. Läs mer om hur det funkar här.
Så här hanterar KB dina uppgifter vid användning av denna tjänst.