High-Level Feature Guided Decoding for Semantic Segmentation
IEEE Transactions on Circuits and Systems for Video Technology · 2024-09 · 34 (9) · 8281-8291
DOI: 10.1109/tcsvt.2024.3393632 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2024-09. Verfasst von Ye Huang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 78 Literaturangaben. Gefördert von: Shenzhen Science and Technology Plan, Shenzhen LongHua Fundamental Research Program.
Abstract (Originalsprache)
Existing pyramid-based upsamplers (e.g. SemanticFPN), although efficient, usually produce less accurate results compared to dilation-based models when using the same backbone. This is partially caused by the contaminated high-level features since they are fused and fine-tuned with noisy low-level features on limited data. To address this issue, we propose to use powerful pre-trained $\boldsymbol {h}$ igh-level $\boldsymbol {f}$ eatures as $\boldsymbol {g}$ uidance (HFG) so that the upsampler can produce robust results. Specifically, only the high-level features from the backbone are used to train the class tokens, which are then reused by the upsampler for classification, guiding the upsampler features to more discriminative backbone features. One crucial design of the HFG is to protect the high-level features from being contaminated by using proper stop-gradient operations so that the backbone does not update according to the noisy gradient from the upsampler. To push the upper limit of HFG, we introduce a $\boldsymbol {c}$ ontext $\boldsymbol {a}$ ugmentation $\boldsymbol {e}$ ncoder (CAE) that can efficiently and effectively operate on the low-resolution high-level feature, resulting in improved representation and thus better guidance. We name our complete solution as the High-Level Feature Guided Decoder (HFGD). We evaluate the proposed HFGD on three benchmarks: Pascal Context, COCOStuff164k, and Cityscapes. HFGD achieves state-of-the-art results among methods that do not use extra training data, demonstrating its effectiveness and generalization ability.
Förderung
- Shenzhen Science and Technology Plan
- Shenzhen LongHua Fundamental Research Program
Zitieren
APA
Huang, Y., Kang, D., Gao, S., Li, W., & Duan, L. (2024). High-Level Feature Guided Decoding for Semantic Segmentation. IEEE Transactions on Circuits and Systems for Video Technology, 34(9), 8281-8291. https://doi.org/10.1109/tcsvt.2024.3393632
Vancouver
Huang Y, Kang D, Gao S, Li W, Duan L. High-Level Feature Guided Decoding for Semantic Segmentation. IEEE Transactions on Circuits and Systems for Video Technology. 2024;34(9):8281-8291. doi:10.1109/tcsvt.2024.3393632
BibTeX
@article{Huang2024,
author = {Huang, Ye and Kang, Di and Gao, Shenghua and Li, Wen and Duan, Lixin},
title = {High-Level Feature Guided Decoding for Semantic Segmentation},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
year = {2024},
volume = {34},
number = {9},
pages = {8281--8291},
doi = {10.1109/tcsvt.2024.3393632}
}
EUVIMED