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 2. Отображено 2.
09-06-2023 дата публикации

Lightweight window pyramid network model and application thereof

Номер: CN116245144A
Автор: ZHOU QUAN, MO YUWEI, JIANG CHENFENG, NI YINGHAO, FAN YAWEN, KANG BIN, ZHANG SUOFEI, WU XIAOFU
Принадлежит:

The invention belongs to the field of image recognition, and particularly relates to a lightweight window pyramid network model and application thereof, and the lightweight window pyramid network model comprises Patchedding, Patchmerging and Transformerblock. The Patchedding module is used for carrying out average division on an input picture, and each obtained block is used as a vector to carry out subsequent attention calculation. The Patchmerging module carries out downsampling on an input feature map, so that the network can carry out feature calculation of different scales, and a plurality of feature maps with different resolutions are obtained. Transformerblock firstly performs attention calculation on windows of different sizes on an input feature map, so that a network can pay attention to features of different scales, then performs lightweight attention calculation on the features, so that interaction of information in different windows is realized, and finally, the features are ...

Подробнее
Номер записи: 1
09-06-2023 дата публикации

Multi-scale hole neighborhood attention calculation backbone network model and application thereof

Номер: CN116246109A
Автор: ZHOU QUAN, JIANG CHENFENG, NI YINGHAO, MO YUWEI, FAN YAWEN, KANG BIN, ZHANG SUOFEI, WU XIAOFU
Принадлежит:

The invention belongs to the field of image recognition, and particularly relates to a multi-scale hole neighborhood attention calculation backbone network model and application thereof, and the model consists of input, image feature extraction and output. The main body part is image feature extraction and has four stages, the first stage comprises an image block embedding layer and a hole Transformer module, and the second stage, the third stage and the fourth stage are the same in structure and comprise a down-sampling layer and a DT module. In the composition of the DT module, a core module is a hole neighborhood attention calculation module which is used for carrying out self-attention calculation on a feature map input into the part, and the global information of the image can be obtained while the local information of the image is obtained through the calculation of the hole neighborhood attention calculation module. Compared with the most advanced backbone network model at present ...

Подробнее
Номер записи: 2