Optimized Detection of Small Objects in Aerial Imagery

View at publisher Summarise with VIRAT
Authors: Bhavya Zaveri, Tejas Gaikwad, Hetansh Sanghavi, Richa Sharma
Cited by: 0
References: 12

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

It is still a challenging task to detect minute objects in the imagery of drones because of poor resolution, changes in scale of objects, occlusion, and background clutters. Our hypothesis in this paper is to present a sophisticated detection framework by using a blend of superior deep learning techniques, such as Slicing Aided Hyper Inference (SAHI), Weighted Box Fusion (WBF), and a hybrid of YOLOv8 and Faster R-CNN. The framework also improves the capability of detecting small and partially obscured targets by using the image slicing method and aggregating the outputs of multiple detectors. The conducted experiments on the VisDrone benchmark have shown that the accuracy, precision and recall of these models are significantly higher than those of regular detection models. The modular structure of the framework also enables easy flexibility to various situations, including city protection, environmental monitoring, and crisis control. On the whole, the current work presents a viable and efficient strategy in solving the long-standing problem of small-object detection in the complicated aerial scenes.

Fields of Study & Subjects

Advanced Neural Network Applications Visual Attention and Saliency Detection Infrastructure Maintenance and Monitoring

Publication details

Journal / source
Biennial International Conference on Nascent Technologies in Engineering (ICNTE)
Pages
1–6
Publication date
16 January 2026
Language
EN