Deformable Convolution and Semi-supervised Learning in Point Clouds for Aneurysm Classification and Segmentation

Hu, Yan; Meijering, Erik; Xia, Yong; Song, Yang

doi:10.1007/978-3-030-92310-5_33

Cited by 8 publications

(13 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Originally proposed for image classification, FixMatch has also been applied to medical image segmentation [27,58]. Similar works tackle instead semantic segmentation in different domains [18,24,78]. The choice of the threshold is however crucial, as it controls the trade-off between the quality of the results and the speed of convergence.…”

Section: Semi-supervised Learning For Image Segmentationmentioning

confidence: 99%

Computational Evaluation of the Combination of Semi-Supervised and Active Learning for Histopathology Image Segmentation with Missing Annotations

Jiménez,

Dierckx,

Amodei

et al. 2023

2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

Real-world segmentation tasks in digital pathology require a great effort from human experts to accurately annotate a sufficiently high number of images. Hence, there is a huge interest in methods that can make use of nonannotated samples, to alleviate the burden on the annotators. In this work, we evaluate two classes of such methods, semi-supervised and active learning, and their combination on a version of the GlaS dataset for gland segmentation in colorectal cancer tissue with missing annotations. Our results show that semi-supervised learning benefits from the combination with active learning and outperforms fully supervised learning on a dataset with missing annotations. However, an active learning procedure alone with a simple selection strategy obtains results of comparable quality.

show abstract

Section: Semi-supervised Learning For Image Segmentationmentioning

confidence: 99%

Computational Evaluation of the Combination of Semi-Supervised and Active Learning for Histopathology Image Segmentation with Missing Annotations

Jiménez,

Dierckx,

Amodei

et al. 2023

2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

show abstract

“…Based on their complementary mechanisms, we propose a simple yet effective pseudo labeling strategy, Cross-Loss Pseudo Labeling (CLP). While the CLP is applied to a popular strategy for semi-supervised semantic segmentation in the literature [11], [14], [38]; pseudo-label from weakly augmented data gives supervision to strongly augmented data for consistency regularization, the key difference is that we have two different decoders for pseudo labeling according to the CE and FL, each decoder gives the loss supervision to the other. Through this approach, we fully exploit the complementary benefits of CE and FL in semi-supervised learning for semantic segmentation.…”

Section: Number Of Available Labelsmentioning

confidence: 99%

“…To address this bias issue, various techniques have been proposed in prior research in the field of semi-supervised semantic segmentation. AEL [14] introduces an adaptive cutmix strategy that enhances consistent learning by weighted random sampling of images for cutmix based on the frequency of class occurrences. This strategy effectively mitigates the bias problem by supplementing consistency for tail classes.…”

Section: Class Bias In Sslmentioning

confidence: 99%

“…In such cases, the predicted logit values have a high variance if a slight perturbation is applied to the input instance, resulting in inconsistent predictions. To address this, methods for consistency regularization to have the same prediction regardless of the intensity of data augmentation are demonstrated to be effective in the literature [11], [14], [38]. Specifically, predicted labels from weakly augmented images A 𝑤 (𝑥 𝑢 𝑖 ) are utilized in the supervision for strongly augmented images The core idea that pseudo labels produced by the FL 𝑞 fl is provided as the supervision for the decoder of CE 𝑔 ce , and likewise 𝑞 ce is fed to the 𝑔 fl , to exploit the complementary properties of two loss functions CE and FL.…”

Section: A Cross-loss Pseudo Labeling (Clp)mentioning

confidence: 99%

See 1 more Smart Citation

Cross-Loss Pseudo Labeling for Semi-Supervised Segmentation

Lee,

Kim,

Heo

2023

IEEE Access

View full text Add to dashboard Cite

Training semantic segmentation models requires pixel-level annotations, leading to a significant labeling cost in dataset creation. To alleviate this issue, recent research has focused on semi-supervised learning, which utilizes only a small amount of annotation. In this context, pseudo labeling techniques are frequently employed to assign labels to unlabeled data based on the model's predictions. However, there are fundamental limitations associated with the widespread application of pseudo labeling in this regard. Since pseudo labels are generally determined by the model's predictions, these labels could be overconfidently assigned even for erroneous predictions, especially when the model has a confirmation bias. We observed that the overconfident prediction tendency of the cross-entropy loss exacerbates this issue, and to address it, we discover the focal loss, known for enabling more reliable confidence estimation, can complement the cross-entropy loss. The cross-entropy loss produces rich labels since it tends to be overconfident. On the other hand, the focal loss provides more conservative confidence, therefore, it produces a smaller number of pseudo labels compared to the cross-entropy. Based on such complementary mechanisms of two loss functions, we propose a simple yet effective pseudo labeling technique, Cross-Loss Pseudo Labeling (CLP), that alleviates the confirmation bias and lack of pseudo label problems. Intuitively, we can mitigate the overconfidence of the cross-entropy with the conservative predictions of the focal loss, while increasing the number of pseudo labels marked by the focal loss based on the cross-entropy. Additionally, CLP also contributes to improving the performance of the tail classes in class-imbalanced datasets through the class bias mitigation effect of the focal loss. In experimental results, our simple CLP improves mIoU by up to +10.4%p compared to a supervised model when only 1/32 true labels are available on PASCAL VOC 2012, and it surpassed the performance of the state-of-the-art methods.

show abstract

“…Although there are many types of equalizers in the literature [6,7], we still need to formulate the coefficients of these. Two main techniques [8], [9] are used to solve this issue: adaptation and automatic synthesis. An automatic synthesis method stores the copy of the input signal as the training signal.…”

Section: Introductionmentioning

confidence: 99%

Performance Analysis of Adaptive Equalizers Over Multipath Faded Channels: Error Vector Magnitudes

Annapurna H. S

2023

IJRITCC

View full text Add to dashboard Cite

Due to the increasing popularity of digital transmission systems, the need for channel equalizers has been acknowledged. These techniques are designed to counteract the effects of the inter-symbol interference (ISI) caused by the communication channels. An adaptive equalizer is used to operate on the output of a channel in order to provide an approximation of the transmission medium. An adaptive equalizer usually requires a training period to operate successfully. This method eliminates the effects of the wireless transmission channel and allows the subsequent symbol modulation. The paper presents an overview of the various adaptive equalizers, such as the least mean squares (LMS), decision feedback equalizers (DFE), and the Recursive least squares (RLS). It also explores their performance in terms of error vector magnitudes (EVM) over Rician and Rayleigh channels. The paper looks into the effects of adaptive equalizers on various digital modulation techniques for rectangular quadrature phase shift, amplitude modulation, such as the BPSK, QPSK, 4-QAM, 16-QAM, 64-QAM and 256-QAM. These modulations are analysed and measured in terms of symbol error rates and number of incorrect symbols.

show abstract

Deformable Convolution and Semi-supervised Learning in Point Clouds for Aneurysm Classification and Segmentation

Cited by 8 publications

References 9 publications

Computational Evaluation of the Combination of Semi-Supervised and Active Learning for Histopathology Image Segmentation with Missing Annotations

Computational Evaluation of the Combination of Semi-Supervised and Active Learning for Histopathology Image Segmentation with Missing Annotations

Cross-Loss Pseudo Labeling for Semi-Supervised Segmentation

Performance Analysis of Adaptive Equalizers Over Multipath Faded Channels: Error Vector Magnitudes

Contact Info

Product

Resources

About