Настройки

Укажите год
-

Небесная энциклопедия

Космические корабли и станции, автоматические КА и методы их проектирования, бортовые комплексы управления, системы и средства жизнеобеспечения, особенности технологии производства ракетно-космических систем

Подробнее
-

Мониторинг СМИ

Мониторинг СМИ и социальных сетей. Сканирование интернета, новостных сайтов, специализированных контентных площадок на базе мессенджеров. Гибкие настройки фильтров и первоначальных источников.

Подробнее

Форма поиска

Поддерживает ввод нескольких поисковых фраз (по одной на строку). При поиске обеспечивает поддержку морфологии русского и английского языка
Ведите корректный номера.
Ведите корректный номера.
Ведите корректный номера.
Ведите корректный номера.
Укажите год
Укажите год
Применить Всего найдено 5. Отображено 5.
02-06-2023 дата публикации

Image classification method based on online knowledge distillation

Номер: CN116206327A
Автор: YANG CHUANGUANG, AN ZHULIN, XU YONGJUN
Принадлежит:

The invention relates to the technical field of image classification, and discloses an image classification method based on online knowledge distillation, which comprises the steps of inputting images to be classified, calibrating portraits in image data, constructing a training image set, extracting global features of the images, superposing local features in the global features, and constructing a plurality of equal network frameworks. The method comprises the steps of constructing an integrated teacher model, aligning probability distribution between the integrated teacher model and equivalent networks, introducing a feature embedding space, obtaining a total comparison loss function under M equivalent networks, introducing a weight factor to obtain a total loss function of a classification model, bringing a data set into a training number set for training, and obtaining an image classification result. The method comprises the following steps: extracting a local feature region, independently ...

Подробнее
Номер записи: 1
23-06-2023 дата публикации

Semantic segmentation knowledge distillation method and system based on cross-image similarity relationship

Номер: CN116310328A
Автор: YANG CHUANGUANG, AN ZHULIN, YU XINQIANG, XU YONGJUN
Принадлежит:

The invention provides a semantic segmentation knowledge distillation method based on a cross-image similarity relationship, and the method comprises the steps: carrying out the statistics of the similarity of pixels between every two feature graphs in a teacher feature graph set, obtaining a teacher image similarity matrix, carrying out the statistics of the similarity of pixels between every two feature graphs in a student feature graph set, and obtaining a student image similarity matrix; and counting the least square error of the teacher image similarity matrix and the student image similarity matrix of the pairwise feature maps as the error of the pairwise feature maps, and collecting the errors of all the pairwise feature maps in the training set to obtain the total error so as to train and update the student model and execute the semantic segmentation task. The student model can learn the global pixel feature relationship dependence of the teacher model by migrating the similarity ...

Подробнее
Номер записи: 2
05-06-2020 дата публикации

Novel road marking structure

Номер: CN0210684512U
Автор: YANG CHUANGUANG, ZHANG AIHUA
Принадлежит: HENAN TIANHUA TRAFFIC TECH CO LTD, HENAN TIANHUA TRAFFIC TECHNOLOGY CO., LTD.

The utility model provides a novel road marking structure which sequentially comprises a primer layer, a hot melt adhesive layer and a reflective layer from bottom to top, reflective glass beads are arranged on the reflective coating, and a noctilucent powder layer is adhered to the surface layers of the reflective glass beads. The road guardrail is simple in structure, convenient to use, capableof obviously improving the safety of a road, firm in structural adhesion of all parts and not prone to falling off, and the service life is prolonged.

Подробнее
Номер записи: 3
18-07-2023 дата публикации

Tilt-rotor aircraft and control method thereof

Номер: CN116443243A
Автор: YANG CHUANGUANG, DU HANG, LIU ZHE, ZHANG FUQIANG, TANG HUSHENG, LYU QUANJIANG
Принадлежит:

The invention belongs to the technical field of aircraft design, and relates to a tilt-rotor aircraft and a control method thereof. Comprising a fuselage, wings, a horizontal tail, a vertical tail and a tiltable coaxial dual-rotor system, the tilting coaxial double rotors are adopted, the rotor disc area is reduced, the tilting rotors can be arranged on the inner sides of the wings to the maximum extent, the two engines are arranged in the middle of the fuselage, the tilting mass is reduced through the measures, the requirement for the structural strength of the wing roots of the wings is lowered, and the structural weight can be reduced easily; meanwhile, the engine is arranged in the middle of the fuselage, so that the problem that a deck or a runway is easily ablated by tail gas exhausted by the engine due to the built-in engine of the wingtip nacelle cloth is solved; the foldable coaxial double rotors are adopted, so that the power requirements of the aircraft in different flight states ...

Подробнее
Номер записи: 4
12-05-2023 дата публикации

Image classification method and system based on multi-model joint contrast learning

Номер: CN116109873A
Автор: YANG CHUANGUANG, AN ZHULIN, YU XINQIANG, XU YONGJUN
Принадлежит:

The invention provides an image classification method and system based on multi-model joint comparative learning, and the method comprises the steps: providing a multi-model joint comparative learning algorithm to guide different models to interact at a feature level, and mutually migrating the respective comparative similarity probability distribution among a plurality of models. By means of comparative learning, samples of the same category can be close to each other in the feature space, samples of different categories can be far away from each other in the feature space, and then a discriminative representation space is obtained. In a scene of multi-model joint learning, comparison learning can enable feature vectors of different models to establish a relationship, so that interaction is carried out on multiple models on a feature level. And through multi-model joint contrast learning, each network can learn better feature expression, so that the image classification effect and the ...

Подробнее
Номер записи: 5