Weakly-supervised semantic segmentation with superpixel guided local and global consistency

Sangbong, Yi; Ma, Huimin; Wang, Xiang; Hu, Tianyu; Li, Xi; Wang, Yu

doi:10.1016/j.patcog.2021.108504

Cited by 23 publications

(4 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [34] , a new set of loss function is designed for weakly supervised labeling, and cross-entropy loss is performed for locally labeled information, and unlabeled regions are evaluated using gated CRF loss. In [35] argues that most research efforts have focused on the localization information ignoring rule-based appearance information. In addition, the article attempts to establish guidance mining semantic affinities between pixels consistent with the image's local and global consistency.…”

Section: Relate Workmentioning

confidence: 99%

Weakly supervised segmentation of COVID-19 infection with local lesion coherence on CT images

Sun

Feng²,

Liu³

et al. 2023

Biomedical Signal Processing and Control

View full text Add to dashboard Cite

Section: Relate Workmentioning

confidence: 99%

Weakly supervised segmentation of COVID-19 infection with local lesion coherence on CT images

Sun

Feng²,

Liu³

et al. 2023

Biomedical Signal Processing and Control

View full text Add to dashboard Cite

“…Fan et al [42] attempted to resolve inaccurate object boundaries by learning the intra-class discriminator under the guidance of superpixels. Yi et al [43] proposed to improve CAM to maintain the local consistency by assigning the same label to pixels that belong to the same superpixel. Unlike these methods that explicitly employ superpixels of the image to improve the quality of CAM, the proposed method performs clustering in the feature space of the attention mechanism to aggregate similar patches, which are further simultaneously attended for generating ClusterCAM.…”

Section: Superpixel Algorithms For Wsssmentioning

confidence: 99%

Clustering-Guided Class Activation for Weakly Supervised Semantic Segmentation

Kim,

Kim

2024

IEEE Access

View full text Add to dashboard Cite

Weakly-supervised semantic segmentation (WSSS) methods via transformer have been actively studied by leveraging their strong capability to capture the global context. However, since the activation function only highlights a few tokens in the self-attention mechanism of the transformer, these methods still suffer from the sparse attention map, which leads to the generation of incomplete pseudo labels. In this paper, we propose a novel class activation scheme that is able to uniformly highlight the whole object region. The key idea of the proposed method is to activate the object region by following the guide of clusters, which are formed by combining similar image features of the object. Specifically, the clustering-guided class activation map (ClusterCAM) is generated from the proposed clustering-based attention module, and highly responsive regions in this map are then adopted to activate target objects in the encoded feature space. This helps the model to explore the entire region of the target object by using the semantic proximity between patch tokens extracted from the same object. Based on this, we design an endto-end WSSS framework that can simultaneously train classification and segmentation networks in a singlestage manner. Experimental results on benchmark datasets show that our proposed method significantly outperforms previous WSSS methods, including several multi-stage approaches. The code and model are publicly available at: https://github.com/DCVL-WSSS/ClusterCAM.

show abstract

“…In order to minimize workload and handle more meaningful data than simple pixels, medical applications [1][2][3], semantic segmentation methods [4], and video segmentation algorithms [5] often recur to superpixel segmentation. In brief, superpixels are disjoint groups of connected pixels that present similar characteristics, like color.…”

Section: Introductionmentioning

confidence: 99%

Novel Arc-Cost Functions and Seed Relevance Estimations for Compact and Accurate Superpixels

et al. 2023

View full text Add to dashboard Cite

On the verge of superpixel methods exploiting saliency information, the Superpixels through Iterative CLEarcutting (SICLE) framework has reported fast and accurate superpixel delineation. It is composed of three steps: (i) seed oversampling; (ii) superpixel generation; and (iii) seed removal. It starts from (i) and applies several iterations of (ii) and (iii) until reaching the desired superpixel quantity. In this work, we improve SICLE such that it can now generate compact superpixels with accurate delineation. We exploit differential computation and propose several novel functions for steps (ii) and (iii) for proper saliency incorporation, compact superpixel generation, and improvement in speed and delineation. Results show that, with our proposals, SICLE achieves state-of-the-art performance in delineation and speed whenever saliency is absent with on-par compacity. When an accurate saliency map is provided, its performance improves significantly and requires only two iterations for segmentation.

show abstract

Weakly-supervised semantic segmentation with superpixel guided local and global consistency

Cited by 23 publications

References 7 publications

Weakly supervised segmentation of COVID-19 infection with local lesion coherence on CT images

Weakly supervised segmentation of COVID-19 infection with local lesion coherence on CT images

Clustering-Guided Class Activation for Weakly Supervised Semantic Segmentation

Novel Arc-Cost Functions and Seed Relevance Estimations for Compact and Accurate Superpixels

Contact Info

Product

Resources

About