It has been difficult to achieve a suitable balance between effectiveness and efficiency in lightweight semantic segmentation networks in recent years. The goal of this work is to present an efficient and reliable semantic segmentation method called EBUNet, which is aimed at achieving a favorable trade-off between inference speed and prediction accuracy. Initially, we develop an Efficient Bottleneck Unit (EBU) that employs depth-wise convolution and depth-wise dilated convolution to obtain adequate features with moderate computation costs. Then, we developed a novel Image Partition Attention Module (IPAM), which divides feature maps into subregions and generates attention weights based on them. As a third step, we developed a novel lightweight attention decoder with which to retrieve spatial information effectively. Extensive experiments show that our EBUNet achieves 73.4% mIou and 152 FPS on the Cityscapes dataset and 72.2% mIoU and 147 FPS on the Camvid dataset with only 1.57 M parameters. The results of the experiment confirm that the proposed model is capable of making a decent trade-off in terms of accuracy, inference, and model size. The source code of our EBUNet is available at (https://github.com/Skybird1101/EBUNet).

The runway detection algorithm for fixed-wing aircraft is a hot topic in the field of aircraft visual navigation. High accuracy, high fault tolerance, and lightweight design are the core requirements in the domain of runway feature detection. This paper aims to address these needs by proposing a lightweight runway feature detection algorithm named YOMO-Runwaynet, designed for edge devices. The algorithm features a lightweight network architecture that follows the YOMO inference framework, combining the advantages of YOLO and MobileNetV3 in feature extraction and operational speed. Firstly, a lightweight attention module is introduced into MnasNet, and the improved MobileNetV3 is employed as the backbone network to enhance the feature extraction efficiency. Then, PANet and SPPnet are incorporated to aggregate the features from multiple effective feature layers. Subsequently, to reduce latency and improve efficiency, YOMO-Runwaynet generates a single optimal prediction for each object, eliminating the need for non-maximum suppression (NMS). Finally, experimental results on embedded devices demonstrate that YOMO-Runwaynet achieves a detection accuracy of over 89.5% on the ATD (Aerovista Runway Dataset), with a pixel error rate of less than 0.003 for runway keypoint detection, and an inference speed exceeding 90.9 FPS. These results indicate that the YOMO-Runwaynet algorithm offers high accuracy and real-time performance, providing effective support for the visual navigation of fixed-wing aircraft.