Search

UAV marine monitoring plays an essential role in marine environmental protection because of its flexibility and convenience, low cost and convenient maintenance. In marine environmental monitoring, the similarity between objects such as oil spill and sea surface, Spartina alterniflora and algae is high, and the effect of the general segmentation algorithm is poor, which brings new challenges to the segmentation of UAV marine images. Panoramic segmentation can do object detection and semantic segmentation at the same time, which can well solve the polymorphism problem of objects in UAV ocean images. Currently, there are few studies on UAV marine image recognition with panoptic segmentation. In addition, there are no publicly available panoptic segmentation datasets for UAV images. In this work, we collect and annotate UAV images to form a panoptic segmentation UAV dataset named UAV-OUC-SEG and propose a panoptic segmentation method named PanopticUAV. First, to deal with the large intraclass variability in scale, deformable convolution and CBAM attention mechanism are employed in the backbone to obtain more accurate features. Second, due to the complexity and diversity of marine images, boundary masks by the Laplacian operator equation from the ground truth are merged into feature maps to improve boundary segmentation precision. Experiments demonstrate the advantages of PanopticUAV beyond the most other advanced approaches on the UAV-OUC-SEG dataset. [ABSTRACT FROM AUTHOR]

Lightweight Network Architecture is essential for autonomous and intelligent monitoring of Unmanned Aerial Vehicles (UAVs), such as in object detection, image segmentation, and crowd counting applications. The state-of-the-art lightweight network learning based on Neural Architecture Search (NAS) usually costs enormous computation resources. Alternatively, low-performance embedded platforms and high-resolution drone images pose a challenge for lightweight network learning. To alleviate this problem, this paper proposes a new lightweight object detection model, called GhostShuffleNet (GSNet), for UAV images, which is built based on Zero-Shot Neural Architecture Search. This paper also introduces the new components which compose GSNet, namely GhostShuffle units (loosely based on ShuffleNetV2) and the backbone GSmodel-L. Firstly, a lightweight search space is constructed with the GhostShuffle (GS) units to reduce the parameters and floating-point operations (FLOPs). Secondly, the parameters, FLOPs, layers, and memory access cost (MAC) as constraints add to search strategy on a Zero-Shot Neural structure search algorithm, which then searches for an optimal network GSmodel-L. Finally, the optimal GSmodel-L is used as the backbone network and a Ghost-PAN feature fusion module and detection heads are added to complete the design of the lightweight object detection network (GSNet). Extensive experiments are conducted on the VisDrone2019 (14.92%mAP) dataset and the our UAV-OUC-DET (8.38%mAP) dataset demonstrating the efficiency and effectiveness of GSNet. The completed code is available at: https://github.com/yfq-yy/GSNet. • Design a new lightweight search space, named GS unit. • Constraints are added on parameters, FLOPs, layers, and MAC to search lighter models. • Construct a new lightweight object detection network, named GSNet. [ABSTRACT FROM AUTHOR]