No abstract

No abstract

Upsampling operation is necessary for semantic segmentation and other pixel-level prediction tasks. Among the commonly used upsampling operations, some are too simple to effectively recover the spatial details lost during downsampling process, and some are too complex and have high computation complexity. In real-world applications, it is critical to achieve high accuracy and maintain real-time inference speed. Therefore, an efficient upsampling operation is essential for these tasks. In this paper, we introduce efficient adaptive upsampling module (EAUM) for real-time semantic segmentation. Inspired by dynamic filter networks, EAUM adaptively predicts the kernel weight of each point in the upsampled feature map according to the corresponding points in the input feature map. To reduce computational cost, EAUM decomposes the spatial information and channel information required for upsampling. The proposed EAUM shows impressive performance on Cityscapes and CamVid benchmarks. Specifically, DenseENet with EAUM outperforms the baseline by 1.4% [Formula: see text] and 1.6% [Formula: see text] in accuracy with a slight drop in inference speed on Cityscapes test dataset.

This paper proposes a method of establishing flight path planning surrogate model based on stacking ensemble learning, which can solve the real-time problem of complex flight mission’s on-line waypoints calculation. Airborne navigation system mainly utilizes several discrete waypoints to guide flight path or flight control. These waypoints are usually calculated by a set of equations based on flight dynamics, flight kinematics and flight mission constraints, and therefore path planning for complex missions cannot guarantee real-time performance. In this paper, flight samples are generated offline by taking flight mission characteristic parameters as input and flight waypoint coordinate series as output. Then two-layer coupling model is constructed based on stacking ensemble learning. A series of base-learners are constructed to learn the quantity of waypoints or each waypoint’s coordinate values respectively. At last, flight path planning surrogate model is built by combining all the base-learners, establishing the direct mapping relationship between input and output. The results show that this surrogate model can effectively calculate the aircraft flight waypoints, and meanwhile maintains ideal accuracy and real-time performance.

No abstract