Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Inhibitory Effects of Isobavachalcone on Tau Protein Aggregation, Tau Phosphorylation, and Oligomeric Tau-Induced Apoptosis.

ACS chemical neuroscience·2020
Same author

Glycyrrhizin, the active compound of the TCM drug Gan Cao stimulates actin remodelling and defence in grapevine.

Plant science : an international journal of experimental plant biology·2020
Same author

His-Rich Domain of Selenoprotein P Ameliorates Neuropathology and Cognitive Deficits by Regulating TrkB Pathway and Zinc Homeostasis in an Alzheimer Model of Mice.

ACS chemical neuroscience·2020
Same author

Corrigendum: Müller Cell Regulated Microglial Activation and Migration in Rats With <i>N</i>-Methyl-<i>N</i>-Nitrosourea-Induced Retinal Degeneration.

Frontiers in neuroscience·2020
Same author

Characterizing glycosyltransferases by a combination of sequencing platforms applied to the leaf tissues of Stevia rebaudiana.

BMC genomics·2020
Same author

ACP-5862 suppresses esophageal squamous cell carcinoma growth through inducing apoptosis via activation of endoplasmic reticulum stress and ROS production.

Biochemical and biophysical research communications·2020

Related Experiment Video

Updated: Nov 3, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

733

Multi-Level and Multi-Scale Feature Aggregation Network for Semantic Segmentation in Vehicle-Mounted Scenes.

Yong Liao1, Qiong Liu1

  • 1School of Software Engineering, South China University of Technology, Guangzhou 510006, China.

Sensors (Basel, Switzerland)
|June 2, 2021
PubMed
Summary
This summary is machine-generated.

We introduce MMFANet, a novel network for real-time semantic segmentation that effectively handles object scale variation. This efficient model achieves high accuracy without compromising speed or increasing computational cost.

Keywords:
feature aggregationmulti-scale feature extractionreal-time semantic segmentationvehicle-mounted scenes

More Related Videos

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

3.1K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

592

Related Experiment Videos

Last Updated: Nov 3, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

733
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

3.1K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

592

Area of Science:

  • Computer Vision
  • Deep Learning
  • Artificial Intelligence

Background:

  • Semantic segmentation in vehicle-mounted scenes faces challenges like object scale variation and the accuracy-efficiency trade-off.
  • Existing lightweight networks often use single-scale features and fixed receptive fields, compromising spatial details for speed.
  • Current real-time segmentation methods often sacrifice accuracy for speed or increase computation with dilated convolutions and sub-networks.

Purpose of the Study:

  • To propose a multi-level and multi-scale feature aggregation network (MMFANet) for efficient and accurate semantic segmentation.
  • To address object scale variation and the accuracy-efficiency trade-off in vehicle-mounted scene segmentation.
  • To develop a network that captures spatial details and high-level semantics without significant computational overhead.

Main Methods:

  • Designed a spatial pyramid module using cascaded dilated convolutions with varying receptive fields for layer-by-layer multi-scale feature extraction.
  • Developed a lightweight backbone network by reducing the feature channel capacity of the spatial pyramid module.
  • Incorporated two additional modules to capture spatial details and high-level semantics, enhancing accuracy with minimal computational increase.

Main Results:

  • Achieved 79.3% MIoU on the Cityscapes test dataset at 58.5 FPS, outperforming SwiftNet (75.5% MIoU).
  • Demonstrated a significant reduction in parameters, with at least 53.38% fewer parameters compared to other state-of-the-art models.
  • Validated the effectiveness of MMFANet in balancing accuracy and efficiency for real-time semantic segmentation.

Conclusions:

  • MMFANet offers a superior solution for real-time semantic segmentation in vehicle-mounted scenes.
  • The proposed network effectively addresses key challenges, providing high accuracy and efficiency.
  • MMFANet presents a computationally efficient alternative with a reduced parameter count for advanced computer vision applications.