YOLO-S: A Lightweight and Accurate YOLO-like Network for Small Target Selection in Aerial Imagery

Small target detection is still a challenging task, especially when looking at fast and accurate solutions for mobile or edge applications. In this work, we present YOLO-S, a simple, fast, and efficient network. It exploits a small feature extractor, as well as skip connection, via both bypass and concatenation, and a reshape-passthrough layer to promote feature reuse across network and combine low-level positional information with more meaningful high-level information. Performances are evaluated on AIRES, a novel dataset acquired in Europe, and VEDAI, benchmarking the proposed YOLO-S architecture with four baselines. We also demonstrate that a transitional learning task over a combined dataset based on DOTAv2 and VEDAI can enhance the overall accuracy with respect to more general features transferred from COCO data. YOLO-S is from 25% to 50% faster than YOLOv3 and only 15-25% slower than Tiny-YOLOv3, outperforming also YOLOv3 by a 15% in terms of accuracy (mAP) on the VEDAI dataset. Simulations on SARD dataset also prove its suitability for search and rescue operations. In addition, YOLO-S has roughly 90% of Tiny-YOLOv3's parameters and one half FLOPs of YOLOv3, making possible the deployment for low-power industrial applications.

Automated summary

In this work, a simple, fast, and efficient network called YOLO-S is proposed for small target detection task. It utilizes a small feature extractor and skip connection, along with a reshape-passthrough layer, to promote feature reuse and combine low-level positional information with high-level information. The performance of YOLO-S is evaluated on AIRES and VEDAI datasets, and it outperforms four baselines in terms of accuracy. The experiments demonstrate that transitional learning on a combined dataset can enhance overall accuracy. YOLO-S is faster than YOLOv3 and only slightly slower than Tiny-YOLOv3, while achieving higher accuracy on VEDAI dataset. It is also suitable for search and rescue operations according to simulations on SARD dataset. Additionally, YOLO-S has a smaller model size and lower computational complexity, making it deployable for low-power industrial applications.