Weakly supervised semantic segmentation (WSSS) models relying on class activation maps (CAMs) have achieved desirable performance comparing to the non-CAMs-based counterparts. However, to guarantee WSSS task feasible, we need to generate pseudo labels by expanding the seeds from CAMs which is complex and time-consuming, thus hindering the design of efficient end-to-end (single-stage) WSSS approaches. To tackle the above dilemma, we resort to the off-the-shelf and readily accessible saliency maps for directly obtaining pseudo labels given the image-level class labels. Nevertheless, the salient regions may contain noisy labels and cannot seamlessly fit the target objects, and saliency maps can only be approximated as pseudo labels for simple images containing single-class objects. As such, the achieved segmentation model with these simple images cannot generalize well to the complex images containing multi-class objects. To this end, we propose an end-to-end multi-granularity denoising and bidirectional alignment (MDBA) model, to alleviate the noisy label and multi-class generalization issues. Specifically, we propose the online noise filtering and progressive noise detection modules to tackle image-level and pixel-level noise, respectively. Moreover, a bidirectional alignment mechanism is proposed to reduce the data distribution gap at both input and output space with simple-to-complex image synthesis and complex-to-simple adversarial learning. MDBA can reach the mIoU of 69.5% and 70.2% on validation and test sets for the PASCAL VOC 2012 dataset. The source codes and models have been made available at https://github.com/NUST-Machine-Intelligence-Laboratory/MDBA.

Download full-text PDF

Source
http://dx.doi.org/10.1109/TIP.2023.3275913DOI Listing

Publication Analysis

Top Keywords

bidirectional alignment
12
pseudo labels
12
multi-granularity denoising
8
denoising bidirectional
8
weakly supervised
8
supervised semantic
8
semantic segmentation
8
saliency maps
8
simple images
8
labels
5

Similar Publications

Research on atrial fibrillation diagnosis in electrocardiograms based on CLA-AF model.

Eur Heart J Digit Health

January 2025

Hunan Key Laboratory of Biomedical Nanomaterials and Devices, Hunan University of Technology, No. 88 West Taishan Road, Zhuzhou 412007, Hunan, China.

Aims: The electrocardiogram (ECG) is the primary method for diagnosing atrial fibrillation (AF), but interpreting ECGs can be time-consuming and labour-intensive, which deserves more exploration.

Methods And Results: We collected ECG data from 6590 patients as YY2023, classified as Normal, AF, and Other. Convolutional Neural Network (CNN), bidirectional Long Short-Term Memory (BiLSTM), and Attention construct the AF recognition model CNN BiLSTM Attention-Atrial Fibrillation (CLA-AF).

View Article and Find Full Text PDF

Many recent approaches to identity share a foundational similarity with ecological psychology, namely, to place identity in its context. That is, they explicitly place identity in its physical and social environments. Yet, we can distinguish at least two different approaches that diverge fundamentally with regards to the role that this "context" has in identity.

View Article and Find Full Text PDF

Introduction: Artificial vascular scaffolds can mimic the structure of natural blood vessels and replace the damaged vessels by implanting them at the injury site to perform the corresponding functions. Electrospinning technology can perfectly combine biological signals and topographical cues to synergistically induce directed cell migration and growth.

Methods: In this study, poly (caprolactone) (PCL) nanofibers, PCL nanofibers uniformly coated with the extracellular matrix derived from endothelial cells (ECd), and bi-directional linear gradient ECd-coated PCL nanofibers were prepared by electrospinning and electrospray techniques to evaluate their effects on the proliferation and migration of Human umbilical vein endothelial cells (HUVECs) and rapid endothelialization.

View Article and Find Full Text PDF

Impact of Eye Contact on Communication during Online Medication Counseling: An Analysis Using the Roter Interaction Analysis System.

Biol Pharm Bull

January 2025

Education Research Center for Clinical Pharmacy, Faculty of Pharmaceutical Sciences, Hokkaido University, Kita 12, Nishi 6, Kita-ku, Sapporo 060-0812, Japan.

We have previously used the Roter Interaction Analysis System (RIAS) to analyze differences between online and face-to-face medication counseling. In our previous research, students have commented that the built-in camera on their laptops makes it difficult to make eye contact and communicate effectively. Furthermore, there is a lack of research on the impact of eye contact in online medical communication.

View Article and Find Full Text PDF

Effective team science requires procedural harmonization for rigor and reproducibility. Multicenter studies across experimental modalities (domains) can help accelerate translation. The Translational Outcomes Project in NeuroTrauma (TOP-NT) is a pre-clinical traumatic brain injury (TBI) consortium charged with establishing and validating noninvasive TBI assessment tools through team science.

View Article and Find Full Text PDF

Want AI Summaries of new PubMed Abstracts delivered to your In-box?

Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!