Small Object Detection in UAV Scenarios Based on YOLOv5

Shuangyuan Li^1,*, Zhengwei Wang², Jiaming Liang³, Yichen Wang⁴
1 Information Center, Jilin University of Chemical Technology, Jilin, 132022, China
2 Chongqing University of Posts and Telecommunications, Chongqing, 400065, China
3 Beijing Chaitin Technology Co., Ltd., Beijing, 100101, China
4 School of Information and Control Engineering, Jilin University of Chemical Technology, Jilin, 132022, China
* Corresponding Author: Shuangyuan Li. Email: email
(This article belongs to the Special Issue: Advanced Image Segmentation and Object Detection: Innovations, Challenges, and Applications)

Computer Modeling in Engineering & Sciences https://doi.org/10.32604/cmes.2025.073896

Received 28 September 2025; Accepted 11 November 2025; Published online 09 December 2025

Download PDF

Abstract

Object detection plays a crucial role in the field of computer vision, and small object detection has long been a challenging issue within this domain. In order to improve the performance of object detection on small targets, this paper proposes an enhanced structure for YOLOv5, termed ATC-YOLOv5. Firstly, a novel structure, AdaptiveTrans, is introduced into YOLOv5 to facilitate efficient communication between the encoder and the detector. Consequently, the network can better address the adaptability challenge posed by objects of different sizes in object detection. Additionally, the paper incorporates the CBAM (Convolutional Block Attention Module) attention mechanism, which dynamically adjusts the weights of different channels in the feature map by introducing a channel attention mechanism. Finally, the paper addresses small object detection by increasing the number of detection heads, specifically designed for detecting high-resolution and minute target objects. Experimental results demonstrate that on the VisDrone2019 dataset, ATC-YOLOv5 outperforms the original YOLOv5, with an improvement in mAP@0.5 from 34.32% to 42.72% and an increase in mAP@[0.5:0.95] from 18.93% to 24.48%.