UN-YOLOv5s: A UAV-Based Aerial Photography Detection Algorithm

Abstract

With the progress of science and technology, artificial intelligence is widely used in various disciplines and has produced amazing results. The research of the target detection algorithm has significantly improved the performance and role of unmanned aerial vehicles (UAVs), and plays an irreplaceable role in preventing forest fires, evacuating crowded people, surveying and rescuing explorers. At this stage, the target detection algorithm deployed in UAVs has been applied to production and life, but making the detection accuracy higher and better adaptability is still the motivation for researchers to continue to study. In aerial images, due to the high shooting height, small size, low resolution and few features, it is difficult to be detected by conventional target detection algorithms. In this paper, the UN-YOLOv5s algorithm can solve the difficult problem of small target detection excellently. The more accurate small target detection (MASD) mechanism is used to greatly improve the detection accuracy of small and medium targets, The multi-scale feature fusion (MCF) path is combined to fuse the semantic information and location information of the image to improve the expression ability of the novel model. The new convolution SimAM residual (CSR) module is introduced to make the network more stable and focused. On the VisDrone dataset, the mean average precision (mAP) of UAV necessity you only look once v5s(UN-YOLOv5s) is 8.4% higher than that of the original algorithm. Compared with the same version, YOLOv5l, the mAP is increased by 2.2%, and the Giga Floating-point Operations Per Second (GFLOPs) is reduced by 65.3%. Compared with the same series of YOLOv3, the mAP is increased by 1.8%, and GFLOPs is reduced by 75.8%. Compared with the same series of YOLOv8s, the detection accuracy of the mAP is improved by 1.1%.

Small Target-YOLOv5: Enhancing the Algorithm for Small Object Detection in Drone Aerial Imagery Based on YOLOv5

“…Therefore, at the same size, their detection precision is lower than the SMT-YOLOv5 model proposed in this study. It is worth mentioning that compared with the models proposed by other scholars (such as KPE-YOLOv5s [ 30 ], UN-YOLOV5s [ 31 ], and FE-YOLOv5s [ 32 ]), SMT-YOLOv5s has a superior performance in terms of mAP@50, mAP@0.5:0.95, and GFLOPS values. KPE-YOLOv5s redesigns the size of the anchor box using the K-Means++ clustering algorithm and introduces the SE attention module, but the network feature fusion part lacks optimization, resulting in relatively poor small object detection precision.…”

“…Anchor Frame Configuration. P2 (3,4), (6,5), (4,8), (11,6) P3 (6,12), (11,11), (10,20), (20,10) P4 (17,18), (31,16), (17,32) In the domain of small target detection in UAV imagery, a significant challenge involves effectively combining multi-scale features [20]. As shown in Figure 2, The original YOLOv5 algorithm used a cascade architecture comprising the feature pyramid network (FPN) [21] and pyramid attention network (PANet) [22] for feature fusion.…”

Real-Time Obstacle Detection with YOLOv8 in a WSN Using UAV Aerial Photography

“…Further, the proposed YOLOv8 exhibits a recall of 78.2%, a precision value of 89.99%, and a mAP of 87.4% when trained with 150 epochs. The testing performance of the proposed obstacle detection based on the YOLOv8 model is compared with the other existing object detection models, i.e., YOLOv7 [28], YOLOv5 [24], SSD [35], and Faster-RCNN [34]. The performance is analyzed for all classes with different iteration steps with our own datasets, as shown in Table 4.…”

“…Another study proposed a YOLOv5-like architecture with ConvMixers and an additional prediction head for object detection using UAVs, which were trained and tested on the VisDrone 2021 dataset. Furthermore, the authors in [24] modified the YOLOv5 architecture for better aerial image analysis performance, concentrating on uses like mapping land usage and environmental monitoring.…”