ORIGINAL RESEARCH article
Front. Plant Sci.
Sec. Sustainable and Intelligent Phytoprotection
This article is part of the Research TopicAdvanced Imaging and Phenotyping for Sustainable Plant Science and Precision Agriculture 4.0View all 11 articles
MDFE-Net: A Multiscale Dilated Feature Enhancement Network for Small Object Detection
Provisionally accepted
- 1Fujian Police College, Fuzhou, China
- 2Fujian Agriculture and Forestry University, Fuzhou, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Due to insufficient feature representation and complex backgrounds, small object detection remains a significant challenge. To solve these problems, this paper proposes a novel small object detection framework named Multiscale Dilated Feature Enhancement Network (MDFE-Net), comprising two innovative plug-and-play modules: Multi-scale Dilated Feature Aggregation (MDFA) and Context Feature Enhancement (CFE). Specifically, the MDFA module efficiently integrates multi-scale features via dilated convolution, capturing rich contextual information and enhancing underlying feature representations. CFE improves the local feature perception and preserves and extracts informative cues for small objects to the greatest extent. Consequently, the proposed network significantly improves small object perception and effectively mitigates interference from complex backgrounds. Comprehensive experiments were conducted on two public datasets (VisDrone and GTSDB) and one self-constructed agricultural dataset (PSD-Node). On the above three datasets, the AP50 of MDFE-Net reached 0.304, 0.952, and 0.895, and the AP is 0.172, 0.805, and 0.476, respectively, which exceeded the benchmark model and the current SOTA method.
Keywords: Context feature, Dilated Convolution, Feature enhancement, Multiscale, Small object detection
Received: 31 Dec 2025; Accepted: 05 Feb 2026.
Copyright: © 2026 Liu, Lin, Zhang, Li and Zhu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence:
Bin Li
Junyan Zhu
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.