Related Experiment Video
Updated: Sep 11, 2025
10:25
Deep Learning-Based Segmentation of Cryo-Electron Tomograms
Published on: November 11, 2022
9.3K
Multiscale feature tuned trans-DeepLabV3+ based semantic segmentation of aerial images using improved red piranha
P Anilkumar1, K Lokesh2, A Naveen Kumar3
1Department of Electronics and Communication Engineering, Mother Theresa Institute of Engineering and Technology, Palamaner, Andhra Pradesh, 517408, India.
Scientific Reports
|August 18, 2025
Summary
This study introduces an effective deep learning method for semantic segmentation of aerial images, addressing limitations in supervised learning with limited data. The new approach enhances image quality and extracts multi-scale features for improved accuracy in remote sensing applications.
Area of Science:
- Remote Sensing
- Computer Vision
- Artificial Intelligence
Background:
- Supervised learning for semantic segmentation of aerial images requires extensive pixel-level labeled data, which is often scarce.
- Existing Deep Semantic Segmentation Networks (DSSN) struggle with efficiency and accuracy when processing high-resolution aerial imagery, especially in extracting multi-scale semantic details.
Purpose of the Study:
- To develop an effective deep learning-based semantic segmentation method for higher-quality aerial images, overcoming the limitations of data scarcity and improving efficiency.
- To enhance the extraction of multi-scale semantic details crucial for accurate segmentation in remote sensing applications.
Main Methods:
- A heuristic technique was employed to design the semantic segmentation method, utilizing standard information sources for aerial photo collection.
- The Multi-Scale RetiNex (MSRN) technique was used for image quality enhancement, followed by the Multiscale Feature Tuned-Trans-Deeplabv3+ (MSTDeepLabV3+) system for feature extraction.
- The Improved Red Piranha Optimization (IRPO) approach was utilized to fine-tune the parameters of the MSTDeepLabV3+ system.
Main Results:
- The proposed method successfully performs semantic segmentation on enhanced aerial images.
- Experimental evaluation demonstrated the excellent performance of the implemented model in terms of accuracy and efficiency for aerial image segmentation.
Conclusions:
- The developed heuristic deep learning technique provides an effective solution for semantic segmentation of aerial images, particularly in scenarios with limited labeled data.
- The combination of MSRN, MSTDeepLabV3+, and IRPO offers a robust framework for improving the accuracy and efficiency of remote sensing image analysis.
