We present Mix3D, a data augmentation technique for segmenting large-scale 3D scenes. Since scene context helps reasoning about object semantics, current works focus on models with large capacity and receptive fields that can fully capture the global context of an input 3D scene. However, strong contextual priors can have detrimental implications like mistaking a pedestrian crossing the street for a car. In this work, we focus on the importance of balancing global scene context and local geometry, with the goal of generalizing beyond the contextual priors in the training set. In particular, we propose a "mixing" technique which creates new training samples by combining two augmented scenes. By doing so, object instances are implicitly placed into novel out-of-context environments, therefore making it harder for models to rely on scene context alone, and instead infer semantics from local structure as well. We perform detailed analysis to understand the importance of global context, local structures and the effect of mixing scenes. In experiments, we show that models trained with Mix3D profit from a significant performance boost on indoor (ScanNet, S3DIS) and outdoor datasets (SemanticKITTI). Mix3D can be trivially used with any existing method, e.g., trained with Mix3D, MinkowskiNet outperforms all prior state-of-the-art methods by a significant margin on the ScanNet test benchmark (78.1% mIoU). Code

Development and improvement of a mathematical model for a large-scale analysis based on the Daubechies discrete wavelet transform will be implemented in an algebraic system possessing a property of ring and field suitable for speech signals processing. Modular codes are widely used in many areas of modern information technologies. The use of these non-positional codes can provide high-speed data processing. Therefore, these algebraic systems should be used in the algorithms of digital processing of signals, which are characterized by processing large amounts of data in real time. In addition, modular codes make it possible to implement large-scale signal processing using the wavelet transform. The paper discusses examples of the Daubechies wavelet transform application. Integer processing, presented in the paper, will reduce the number of rounding errors when processing the speech signals.

The paper focuses on the new trend of increasing the accuracy of low altitudes measurement by frequency-modulated continuous-wave (FMCW) radio altimeters. The method of increasing the altitude measurement accuracy has been realized in a form of a frequency deviation increase with the help of the carrier frequency increase. In this way, the height measurement precision has been established at the value of ±0.75 m. Modern digital processing of a differential frequency cannot increase the accuracy limitation considerably. It can be seen that further increase of the height measurement precision is possible through the method of innovatory processing of so-called height pulses. This paper thoroughly analyzes the laws of height pulse shaping from the differential frequency presented by the number that represents the information about the measured altitude for this purpose. This paper presents the results of the laboratory experimental altitude measurement with the use of a so-called double-channel method. The application of obtained results could contribute to the increase of air traffic safety, mainly in the phase of the aircraft approaching for landing and landing itself.

No abstract