Threshold Matters in WSSS: Manipulating the Activation for the Robust and Accurate Segmentation Model Against Thresholds

Cited 15 time in webofscience Cited 0 time in scopus
  • Hit : 75
  • Download : 0
DC FieldValueLanguage
dc.contributor.authorLee, Minhyunko
dc.contributor.authorKim, Dongseobko
dc.contributor.authorShim, Hyunjungko
dc.date.accessioned2023-03-17T06:00:12Z-
dc.date.available2023-03-17T06:00:12Z-
dc.date.created2023-03-17-
dc.date.issued2022-06-22-
dc.identifier.citation2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, pp.4320 - 4329-
dc.identifier.issn1063-6919-
dc.identifier.urihttp://hdl.handle.net/10203/305670-
dc.description.abstractWeakly-supervised semantic segmentation (WSSS) has recently gained much attention for its promise to train segmentation models only with image-level labels. Existing WSSS methods commonly argue that the sparse coverage of CAM incurs the performance bottleneck of WSSS. This paper provides analytical and empirical evidence that the actual bottleneck may not be sparse coverage but a global thresholding scheme applied after CAM. Then, we show that this issue can be mitigated by satisfying two conditions; 1) reducing the imbalance in the foreground activation and 2) increasing the gap between the foreground and the background activation. Based on these findings, we propose a novel activation manipulation network with a per-pixel classification loss and a label conditioning module. Per-pixel classification naturally induces two-level activation in activation maps, which can penalize the most discriminative parts, promote the less discriminative parts, and deactivate the background regions. Label conditioning imposes that the output label of pseudo-masks should be any of true image-level labels; it penalizes the wrong activation assigned to non-target classes. Based on extensive analysis and evaluations, we demonstrate that each component helps produce accurate pseudo-masks, achieving the robustness against the choice of the global threshold. Finally, our model achieves state-of-the-art records on both PAS-CAL VOC 2012 and MS COCO 2014 datasets. The code is available at https://github.com/gaviotas/AMN.-
dc.languageEnglish-
dc.publisherIEEE Computer Society-
dc.titleThreshold Matters in WSSS: Manipulating the Activation for the Robust and Accurate Segmentation Model Against Thresholds-
dc.typeConference-
dc.identifier.wosid000867754204057-
dc.identifier.scopusid2-s2.0-85141815600-
dc.type.rimsCONF-
dc.citation.beginningpage4320-
dc.citation.endingpage4329-
dc.citation.publicationname2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022-
dc.identifier.conferencecountryUS-
dc.identifier.conferencelocationNew Orleans-
dc.identifier.doi10.1109/CVPR52688.2022.00429-
dc.contributor.localauthorShim, Hyunjung-
dc.contributor.nonIdAuthorLee, Minhyun-
dc.contributor.nonIdAuthorKim, Dongseob-
Appears in Collection
AI-Conference Papers(학술대회논문)
Files in This Item
There are no files associated with this item.
This item is cited by other documents in WoS
⊙ Detail Information in WoSⓡ Click to see webofscience_button
⊙ Cited 15 items in WoS Click to see citing articles in records_button

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0