Weakly Supervised Semantic Segmentation for Joint Key Local Structure Localization and Classification of Aurora Image

Weakly supervised semantic segmentation (WSSS) has gained significant popularity as it relies only on weak labels such as image level annotations rather than the pixel level annotations required by supervised semantic segmentation (SSS) methods. Despite drastically reduced annotation costs, typical feature representations learned from WSSS are only representative of some salient parts of objects and less reliable compared to SSS due to the weak guidance during training. In this paper, we propose a novel Multi-Strategy Contrastive Learning (MuSCLe) framework to obtain enhanced feature representations and improve WSSS performance by exploiting similarity and dissimilarity of contrastive sample pairs at image, region, pixel and object boundary levels. Extensive experiments demonstrate the effectiveness of our method and show that MuSCLe outperforms current state-of-the-art methods on the widely used PASCAL VOC 2012 dataset.

Recently proposed methods for weakly-supervised semantic segmentation have achieved impressive performance in predicting pixel classes despite being trained with only image labels which lack positional information. Because image annotations are cheaper and quicker to generate, weak supervision is more practical than full supervision for training segmentation algorithms. These methods have been predominantly developed to solve the background separation and partial segmentation problems presented by natural scene images and it is unclear whether they can be simply transferred to other domains with different characteristics, such as histopathology and satellite images, and still perform well. This paper evaluates state-of-the-art weakly-supervised semantic segmentation methods on natural scene, histopathology, and satellite image datasets and analyzes how to determine which method is most suitable for a given dataset. Our experiments indicate that histopathology and satellite images present a different set of problems for weakly-supervised semantic segmentation than natural scene images, such as ambiguous boundaries and class co-occurrence. Methods perform well for datasets they were developed on, but tend to perform poorly on other datasets. We present some practical techniques for these methods on unseen datasets and argue that more work is needed for a generalizable approach to weakly-supervised semantic segmentation. Our full code implementation is available on GitHub: https://github.com/lyndonchan/wsss-analysis.

Existing weakly supervised semantic segmentation (WSSS) methods based on image‐level labels always rely on class activation maps (CAMs), which measure the relationships between features and classifiers. However, CAMs only focus on the most discriminative regions of images, resulting in their poor coverage performance. We attribute this to the deficiency in the recognition ability of a single classifier and the negative impacts caused by magnitudes during the CAMs normalisation process. To address the aforementioned issues, we propose to construct selective multiple classifiers (SMC). During the training process, we extract multiple prototypes for each class and store them in the corresponding memory bank. These prototypes are divided into foreground and background prototypes, with the former used to identify foreground objects and the latter aimed at preventing the false activation of background pixels. As for the inference stage, multiple prototypes are adaptively selected from the memory bank for each image as SMC. Subsequently, CAMs are generated by measuring the angle between SMC and features. We enhance the recognition ability of classifiers by adaptively constructing multiple classifiers for each image, while only relying on angle measurement to generate CAMs can alleviate the suppression phenomenon caused by magnitudes. Furthermore, SMC can be integrated into other WSSS approaches to help generate better CAMs. Extensive experiments conducted on standard WSSS benchmarks such as PASCAL VOC 2012 and MS COCO 2014 demonstrate the superiority of our proposed method.

Pixel-level annotation of lung cavities (LCs) in computed tomography (CT) images is challenging due to their morphological diversity and complexity. Weakly supervised semantic segmentation (WSSS) methods, which utilize sparse annotations (e.g., image-level labels), offer a promising solution. However, existing WSSS approaches often generate coarse pseudo-labels and lack sufficient spatial supervision, resulting in under- or over-segmentation of irregular lesions. To address these limitations, we introduce several key innovations. First, we propose a novel Graph-based Affinity Network (GA-Net) that, unlike conventional methods relying on low-level pixel features, models long-range contextual relationships and structural dependencies using a superpixel graph and learned edge inference kernel, enabling structure-aware pseudo-label refinement for complex lesion morphology. Second, we introduce region-wise affinity propagation, which refines segmentation by propagating activations within semantically coherent 3D regions, offering more precise control over under-/over-segmentation compared to global affinity methods. Additionally, we incorporate Exponential Moving Average (EMA) ensembling for training stability and a scribble-based segmentation module that utilizes pseudo-label contours to provide direct boundary supervision. Extensive experiments on three benchmark datasets demonstrate that our method outperforms existing state-of-the-art medical WSSS techniques, achieving precise and reliable segmentation of complex LCs in CT scans.

Significant advances in weakly-supervised semantic segmentation (WSSS) methods with image-level labels have been made, but they have several key limitations: incomplete object regions, object boundary mismatch, and co-occurring pixels from non-target objects. To address these issues, we propose a novel joint learning framework, namely Saliency Map and Visual Word Encoder (SMVWE), which employs two weak supervisions to generate the high-quality pseudo labels. Specifically, we develop a visual word encoder to encode the localization map into semantic words with a learnable codebook, making the network generate localization maps containing more semantic regions with the encoded fine-grained semantic words. Moreover, to obtain accurate object boundaries and eliminate co-occurring pixels, we design a saliency map selection mechanism with the pseudo-pixel feedback to separate the foreground from the background. During joint learning, we fully utilize the cooperation relationship between semantic word labels and saliency maps to generate high-quality pseudo-labels, thus remarkably improving the segmentation accuracy. Extensive experiments demonstrate that our proposed method better tackles above key challenges of WSSS and obtains the state-of-the-art performance on the PASCAL VOC 2012 segmentation benchmark.

It is generally accepted that one of the critical parts of current vision algorithms based on deep learning and convolutional neural networks is the annotation of a sufficient number of images to achieve competitive performance. This is particularly difficult for semantic segmentation tasks since the annotation must be ideally generated at the pixel level. Weakly-supervised semantic segmentation aims at reducing this cost by employing simpler annotations that, hence, are easier, cheaper and quicker to produce. In this paper, we propose and assess a new weakly-supervised semantic segmentation approach making use of a novel loss function whose goal is to counteract the effects of weak annotations. To this end, this loss function comprises several terms based on partial cross-entropy losses, being one of them the Centroid Loss. This term induces a clustering of the image pixels in the object classes under consideration, whose aim is to improve the training of the segmentation network by guiding the optimization. The performance of the approach is evaluated against datasets from two different industry-related case studies: while one involves the detection of instances of a number of different object classes in the context of a quality control application, the other stems from the visual inspection domain and deals with the localization of images areas whose pixels correspond to scene surface points affected by a specific sort of defect. The detection results that are reported for both cases show that, despite the differences among them and the particular challenges, the use of weak annotations do not prevent from achieving a competitive performance level for both.

The costly process of obtaining semantic segmentation labels has driven research towards weakly supervised semantic segmentation (WSSS) methods, using only image-level, point, or box labels. Such annotations introduce limitations and challenges that results in overly-tuned methods specialized in specific domains or scene types. The over reliance of image-level based methods on generation of high quality class activation maps (CAMs) results in limited applicable dataset complexity range, mostly focusing on object centric scenes. Additionally, the lack of dense annotations requires methods to increase network complexity to obtain additional semantic information, often done through multiple stages of training and refinement. Here, we present a single-stage approach generalizable to a wide range of dataset complexities, that is trainable from scratch, without any dependency on pre-trained backbones, classification, or separate refinement tasks. We utilize point annotations to generate reliable, on-the-fly pseudo-masks through refined and spatially filtered features. We are to demonstrate SOTA performance on benchmark datasets (PascalVOC 2012), as well as significantly outperform other SOTA WSSS methods on recent real-world datasets (CRAID, CityPersons, IAD, ADE20K, CityScapes) with up to 28.1% and 22.6% performance boosts compared to our single-stage and multi-stage baselines respectively.

Precise tumor localization and sub-region identification are critical for disease diagnosis. However, current Weakly Supervised Semantic Segmentation (WSSS) methods for brain tumor segmentation are primarily single-class, neglecting differences between tumor sub-regions. We observed that when mainstream transformer-based WSSS methods are applied to multi-class brain tumor segmentation, they encounter two major challenges: sub-region discrimination errors and over-segmentation of small lesions. To address these challenges and advance multi-class WSSS methods for brain tumor analysis, this paper proposes Channel-gated Transformers with Affinity CAM (CTAC). CTAC first employs channel-gated multi-head self-attention to overcome the over-smoothing tendency of the transformer, thereby enhancing inter-class discriminability and improving the model's subclass differentiation capability. Then, CTAC uses multi-scale smoothed affinity to adaptively suppress low-confidence responses in the Class Activation Map (CAM), mitigating over-activation in the CAM, and alleviating the over-segmentation phenomena of small lesions. The proposed CTAC significantly outperformed the baseline method on the BraTS2021 glioma and BraTS2023-MEN meningioma datasets. On Brats2021, it achieved a multi-class mean IoU (mIoU) of 61.718%, an increase of 4.964 percentage points (pp), with the whole-tumor mIoU reaching 79.798% (+6.882 pp). On Brats2023-MEN, CTAC attained 72.887% mIoU (+4.676 pp) for multi-class segmentation and 75.394% (+7.839 pp) for whole-tumor. Furthermore, CTAC surpasses recent state-of-the-art methods. Code is available at https://github.com/yhan94-lab/CTAC.

Weakly supervised semantic segmentation (WSSS) methods are often built on pixel-level localization maps obtained from a classifier. However, training on class labels only, classifiers suffer from the spurious correlation between fore-ground and background cues (e.g. train and rail), fundamentally bounding the performance of WSSS. There have been previous endeavors to address this issue with additional supervision. We propose a novel source of information to distinguish foreground from the background: Out-of-Distribution (OoD) data, or images devoid of foreground object classes. In particular, we utilize the hard OoDs that the classifier is likely to make false-positive predictions. These samples typically carry key visual features on the background (e.g. rail) that the classifiers often confuse as foreground (e.g. train), so these cues let classifiers correctly suppress spurious background cues. Acquiring such hard OoDs does not require an extensive amount of annotation efforts; it only incurs a few additional image-level labeling costs on top of the original efforts to collect class labels. We propose a method, W-OoD, for utilizing the hard OoDs. W-OoD achieves state-of-the-art performance on Pascal VOC 2012. The code is available at: https://github.com/naver-ai/w-ood.

Image fusion at pixel level without precise registration always causes pseudo colors and other problem. Classification-based fusion scheme can effectively eliminate the false color at the edge of objective. However, the traditional per-pixel classification results in the well-known salt and pepper effect. The only way to smooth the image is to use filters, while impacted on the result of fusion. This paper proposes a method consist of a sequential application of segmentation, classification and fusion techniques. First, the image was multi-resolutely segmented into homogenous areas, and classified it by using the membership functions classifier and additional empirical rules. Subsequently, according to the restriction of the precise classification result, adjusting the multi-spectral image then achieved the fusion by using HSV color transformation. Finally, after compared the statistical properties of the fusion result by different methods, the proposed method showed satisfied result. I. INTRODUCTION With the appearance of plentiful kinds of remote sensing imagery, fusion of satellite images of different spatial resolutions play an important role. Via image fusion, we can obtain more information than can be derived from single kind of image. Pohl prompted that according to the stage which image fusion is performed, there are pixel, feature and decision level. Image fusion at pixel level always was effected by the accuracy of registration, the number of mixed pixel and other factors. The pixel in high-resolution satellite image is smaller sizes and combined with fewer spectral bands that cause greater spectral variation within a class and a greater degree of shadow. Hence, color break and pseudo colors often occur at the edge of object and shadow which fusing the high-resolution image only at pixel levels. During the high-resolution image fusion, for eliminating color break and pseudo colors, using its classification image as the prior knowledge and restricting the fusion area can take full advantage of spectral information from multi-spectral bands and texture information come from pan band to realize the image fusion. It's a method of image fusion that performing at pixel level and integrate with feature level fusion. In this way, phenomenon of pseudo colors can be avoided. In this paper, we propose an image fusion method based on object-oriented classification. Via image segmentation, feature extraction and image fusion based on object-oriented classification, the fusion image can be achieved. In this method, object-oriented classification more appropriate for high-resolution image to get accuracy classification result and overcoming the limit of per-pixel image analyses. Because much information is acquired in the relationship between adjacent pixels, including texture and shape information, which allows for identification of individual objects as oppose to single pixels

Recent work has demonstrated the Transformer model is effective for computer vision tasks. However, the global self-attention mechanism utilized in Transformer models does not adequately consider the local structure and details of images, which may result in the loss of information and local details, causing decreased estimation accuracy in gaze estimation tasks when compared to convolution or sequential stacking methods. To address this issue, we propose a parallel CNNs-Transformer aggregation network (CTA-Net) for gaze estimation, which fully leverages the advantages of the Transformer model in modeling global context while the convolutional neural networks (CNNs) model in retaining local details. Specifically, Transformer and ResNet are deployed to extract facial and eye information, respectively. Additionally, an attention cross fusion (ACFusion) Block is embedded with CNN branch, which decomposes features in space and channels to supplement lost features, suppress noise, and help extract eye features more effectively. Finally, a dual-feature aggregation (DFA) module is proposed to effectively fuse the output features of both branches with the help feature a selection mechanism and a residual structure. Experimental results on the MPIIGaze and Gaze360 datasets demonstrate that our CTA-Net achieves state-of-the-art results.

We represent local spatial structure in a color image using feature matrices that are computed from an image region. Feature matrices contain significantly more information about local image structure than previous representations. Although feature matrices are useful for surface recognition, this representation depends on the spectral properties of the scene illumination. Using a finite dimensional linear model for surface spectral reflectance with the same number of parameters as the number of color bands, we show that illumination changes correspond to linear transformations of the feature matrices and that surface rotations correspond to circular shifts of the matrices. From these relationships we derive an algorithm for illumination and geometry invariant recognition of local surface structure. We demonstrate the algorithm with a series of experiments on images of real objects.

Weakly Supervised Semantic Segmentation (WSSS) has received increasing attentions recently, which leverages only limited supervision to train the segmentation networks. Meanwhile, Transformer utilizes multi-head attention mechanism to overcome the limited receptive fields of convolution neural networks (CNN). As a result, a large amount of WSSS work has been built upon Transformer architecture. However, most of them overlook over-smoothing issue (i.e. the patch tokens in deep layers tend to be similar, indicating that the learned information is nearly identical), which will affect the accuracy of Class Activation Map (CAM) in representing the pixel level semantic information. In this paper, an Attention-based Layer Fusion scheme with Token Masking is proposed to address the above issue. The core modules include a Weighted Layer Attention Aggregation (WLA) module and a Random-Masked Class Token Refinement (RMTR) module. The former integrates the attention information of different layers in an adaptive way, which refines CAM to obtain a refined object contour. The latter introduces random masks to aggregate higher-level semantic information. Extensive experiments have been conducted on public datasets, in which our network manifests superior performances than other state-of-the-arts methods.

Superpixel conditional generation adversarial network for CMR artifact correction

Different kinds of local image structures (such as homogeneous, edge-like and junction-like patches) can be distinguished by the intrinsic dimensionality of the local signals. Intrinsic dimensionality makes use of variance from a point and a line in spectral representation of the signal in order to classify it as homogeneous, edge-like or junction-like. The concept of intrinsic dimensionality has been mostly exercised using discrete formulations; however, recent work (; ) has introduced a continuous definition. The current study analyzes the distribution of local patches in natural images according to this continuous understanding of intrinsic dimensionality. This distribution reveals specific patterns than can be also associated to local image structures established in computer vision and which can be related to orientation and optic flow features. In particular, we link quantitative and qualitative properties of optic-flow error estimates to these patterns. In this way, we also introduce a new tool for better analysis of optic flow algorithms.