SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed Detection
The limited processing capacity of computing equipment that is usually mounted on unmanned aerial vehicles (UAVs) often prevents real-time execution of computer vision tasks, such as image segmentation. In this article, we introduce SqueezeSlimU-Net (SSU-Net), an adaptive and efficient deep learning...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10857312/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850189367588421632 |
|---|---|
| author | Alina L. Machidon Andraz Krasovec Veljko Pejovic Octavian M. Machidon |
| author_facet | Alina L. Machidon Andraz Krasovec Veljko Pejovic Octavian M. Machidon |
| author_sort | Alina L. Machidon |
| collection | DOAJ |
| description | The limited processing capacity of computing equipment that is usually mounted on unmanned aerial vehicles (UAVs) often prevents real-time execution of computer vision tasks, such as image segmentation. In this article, we introduce SqueezeSlimU-Net (SSU-Net), an adaptive and efficient deep learning (DL) model designed to enhance UAV capabilities in performing complex image segmentation tasks under resource constraints, thereby advancing real-time UAV vision—a crucial technology in fields, such as precision agriculture. SSU-Net combines benefits of three specialized DL architectures: the semantic segmentation capabilities of the U-Net architecture, the computational efficiency of SqueezeNet's fire modules, and the dynamic adaptability of slimmable neural networks. This integration allows SSU-Net to adjust its network width in real-time, thus striking the balance between inference accuracy and computational load based on the operational parameters such as task requirements and UAV's battery life. To validate SSU-Net's efficacy, we applied it to a weed detection task using two UAV-collected datasets and tested it on an edge computing platform for UAVs. Our experiments show that SSU-Net can reduce inference energy consumption by up to 65% with only a minimal 2% reduction in accuracy. A comparative evaluation with other state-of-the-art DL image segmentation approaches shows that SSU-Net achieves on par weed detection performance while requiring significantly fewer model parameters. In addition, SSU-Net outperforms state-of-the-art network pruning techniques in balancing accuracy and resource usage. Timing benchmarks show SSU-Net fostering real-time weed detection even on low-resource UAVs, making it ideal for UAV remote sensing applications. |
| format | Article |
| id | doaj-art-ce631c4332994ec5acf10bef3fba2413 |
| institution | OA Journals |
| issn | 1939-1404 2151-1535 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj-art-ce631c4332994ec5acf10bef3fba24132025-08-20T02:15:38ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01185749576410.1109/JSTARS.2025.353617510857312SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed DetectionAlina L. Machidon0https://orcid.org/0000-0002-9330-3865Andraz Krasovec1https://orcid.org/0009-0007-4077-0826Veljko Pejovic2https://orcid.org/0000-0002-9009-0024Octavian M. Machidon3https://orcid.org/0000-0003-3133-1008Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, SloveniaFaculty of Computer and Information Science, University of Ljubljana, Ljubljana, SloveniaFaculty of Computer and Information Science, University of Ljubljana, Ljubljana, SloveniaFaculty of Computer and Information Science, University of Ljubljana, Ljubljana, SloveniaThe limited processing capacity of computing equipment that is usually mounted on unmanned aerial vehicles (UAVs) often prevents real-time execution of computer vision tasks, such as image segmentation. In this article, we introduce SqueezeSlimU-Net (SSU-Net), an adaptive and efficient deep learning (DL) model designed to enhance UAV capabilities in performing complex image segmentation tasks under resource constraints, thereby advancing real-time UAV vision—a crucial technology in fields, such as precision agriculture. SSU-Net combines benefits of three specialized DL architectures: the semantic segmentation capabilities of the U-Net architecture, the computational efficiency of SqueezeNet's fire modules, and the dynamic adaptability of slimmable neural networks. This integration allows SSU-Net to adjust its network width in real-time, thus striking the balance between inference accuracy and computational load based on the operational parameters such as task requirements and UAV's battery life. To validate SSU-Net's efficacy, we applied it to a weed detection task using two UAV-collected datasets and tested it on an edge computing platform for UAVs. Our experiments show that SSU-Net can reduce inference energy consumption by up to 65% with only a minimal 2% reduction in accuracy. A comparative evaluation with other state-of-the-art DL image segmentation approaches shows that SSU-Net achieves on par weed detection performance while requiring significantly fewer model parameters. In addition, SSU-Net outperforms state-of-the-art network pruning techniques in balancing accuracy and resource usage. Timing benchmarks show SSU-Net fostering real-time weed detection even on low-resource UAVs, making it ideal for UAV remote sensing applications.https://ieeexplore.ieee.org/document/10857312/Adaptive neural networkscomputational efficiencyimage segmentationprecision agriculturereal-time unmanned aerial vehicle (UAV) visionweed detection |
| spellingShingle | Alina L. Machidon Andraz Krasovec Veljko Pejovic Octavian M. Machidon SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed Detection IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Adaptive neural networks computational efficiency image segmentation precision agriculture real-time unmanned aerial vehicle (UAV) vision weed detection |
| title | SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed Detection |
| title_full | SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed Detection |
| title_fullStr | SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed Detection |
| title_full_unstemmed | SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed Detection |
| title_short | SqueezeSlimU-Net: An Adaptive and Efficient Segmentation Architecture for Real-Time UAV Weed Detection |
| title_sort | squeezeslimu net an adaptive and efficient segmentation architecture for real time uav weed detection |
| topic | Adaptive neural networks computational efficiency image segmentation precision agriculture real-time unmanned aerial vehicle (UAV) vision weed detection |
| url | https://ieeexplore.ieee.org/document/10857312/ |
| work_keys_str_mv | AT alinalmachidon squeezeslimunetanadaptiveandefficientsegmentationarchitectureforrealtimeuavweeddetection AT andrazkrasovec squeezeslimunetanadaptiveandefficientsegmentationarchitectureforrealtimeuavweeddetection AT veljkopejovic squeezeslimunetanadaptiveandefficientsegmentationarchitectureforrealtimeuavweeddetection AT octavianmmachidon squeezeslimunetanadaptiveandefficientsegmentationarchitectureforrealtimeuavweeddetection |