Efficient-Lightweight YOLO: Improving Small Object Detection in YOLO for Aerial Images

Abstract

The most significant technical challenges of current aerial image object-detection tasks are the extremely low accuracy for detecting small objects that are densely distributed within a scene and the lack of semantic information. Moreover, existing detectors with large parameter scales are unsuitable for aerial image object-detection scenarios oriented toward low-end GPUs. To address this technical challenge, we propose efficient-lightweight You Only Look Once (EL-YOLO), an innovative model that overcomes the limitations of existing detectors and low-end GPU orientation. EL-YOLO surpasses the baseline models in three key areas. Firstly, we design and scrutinize three model architectures to intensify the model’s focus on small objects and identify the most effective network structure. Secondly, we design efficient spatial pyramid pooling (ESPP) to augment the representation of small-object features in aerial images. Lastly, we introduce the alpha-complete intersection over union (α-CIoU) loss function to tackle the imbalance between positive and negative samples in aerial images. Our proposed EL-YOLO method demonstrates a strong generalization and robustness for the small-object detection problem in aerial images. The experimental results show that, with the model parameters maintained below 10 M while the input image size was unified at 640 × 640 pixels, the APS of the EL-YOLOv5 reached 10.8% and 10.7% and enhanced the APs by 1.9% and 2.2% compared to YOLOv5 on two challenging aerial image datasets, DIOR and VisDrone, respectively.

AIOD-YOLO: an algorithm for object detection in low-altitude aerial images

“…26 However, large kernel convolutions also unavoidably increase the model's computational complexity. To address the problem of model lightening in aerial images, EL-YOLO 27 solves the problem of positive and negative sample imbalance by improving the loss function. Deng et al 28 improved YOLOv5s using a lightened backbone and proposed a lightened model LAI-YOLOv5s.…”

SODCNN: A Convolutional Neural Network Model for Small Object Detection in Drone-Captured Images

“…It can be seen that the performance of our algorithm improves more significantly on small targets. We conducted comparative experiments involving prominent object detection algorithms on the VisDrone dataset, including YOLOv3 [44], YOLOv4 [45], YOLOv5l, YOLOv6s [46], YOLOv8s [47], Cascade R-CNN, RetinaNet, TPH-YOLOv5, PicoDet [48], PP-YOLOE [49], and EL-YOLOv5s [50]. By referring to the data presented in Table 8, it is evident that our proposed algorithm surpasses these models in terms of performance.…”

Post-secondary classroom teaching quality evaluation using small object detection model

“…Small target detection techniques 21 , 22 are an important research direction in computer vision aimed at addressing the problem of detecting and locating small objects in images or videos. Several optimization methods 23 , 24 based on existing object detection algorithms have been proposed to reduce the cases of missed detection and false detection for small targets, thereby improving the detection performance of small targets. Research has shown that data augmentation 25 , 26 can improve the detection performance of small targets by addressing the issues of low resolution, limited dataset quantity, and uneven distribution.…”