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 10. Отображено 10.
09-06-2023 дата публикации

Pain expression assessment method based on multi-task transformer

Номер: CN116246326A
Автор: MAO SHASHA, LI ANGZE, HE JINGJIE, GOU SHUIPING, JIAO LICHENG
Принадлежит:

The invention relates to a pain expression assessment method based on a multi-task transformer, and belongs to the technical field of image processing and mode recognition, and the method comprises the following steps: S1, preprocessing expression data; s2, a multi-task transform pain assessment model is established, and S2, a multi-task transform pain assessment model is established; s3, acquiring an input image; s4, performing local feature extraction on the input image; s5, performing weighted combination based on the local features of the self-attention; s6, classifying pain expressions; s7, performing pain expression score regression; s8, updating network parameters; and S9, repeating the steps S3-S8 until the loss function value tends to be stable. According to the method, the relationship between local features is further mined by using a transform structure, the quality of the extracted features is improved, a classification and regression multi-task model is adopted, the problem ...

Подробнее
Номер записи: 1
18-07-2023 дата публикации

Class increment multi-organ segmentation method based on general and private feature joint domain representation

Номер: CN116452515A
Автор: TONG NUO, GOU SHUIPING, WANG KUNRU, LI JUANFEI, HUANG LUGUANG, YAO YAO, MENG QINGYANG
Принадлежит:

The invention discloses a class increment multi-organ segmentation method based on general and private feature joint domain representation, and mainly solves the problem of low segmentation precision caused by high specificity of an anatomical region and a segmentation target in an existing method. According to the implementation scheme, a multi-anatomical-region image data set and an abdomen CT image data set are obtained and preprocessed, and the data sets are divided into a training set and a test set; using the constructed ISRNet network to train a general representation model; constructing a class increment multi-organ segmentation model based on general and private feature joint domain representation; carrying out iterative training on the model by utilizing an abdominal CT image segmentation training set through a back propagation method; and inputting the test set into the trained class increment multi-organ segmentation model, and outputting a segmentation result. According to ...

Подробнее
Номер записи: 2
14-07-2023 дата публикации

Feature adversarial network-based heterogeneous image template matching method

Номер: CN116433735A
Автор: GUO ZHANG, LI SHENG, ZHENG ZHIHUI, GOU SHUIPING, WANG XINLIN, MAO SHASHA
Принадлежит:

The invention discloses a heterogeneous image template matching method based on a feature adversarial network. The method comprises the following steps: constructing a training set; respectively constructing a feature extraction network, a feature mapping layer and a modal discrimination network; and carrying out iterative training on the network by using a generative adversarial strategy by utilizing the training data, and matching different-source images of the to-be-matched data set. According to the method, based on adversarial learning, a maximum-minimum game is carried out through a feature mapper and a modal discrimination network to achieve dynamic Nash equilibrium. In a feature mapper, texture complexity calculation and matching result constraint are introduced to ensure that cross-modal features still have good characterization capability after being projected to a common subspace besides calculation of structural loss of mapping features of two modals, so that homologous conversion ...

Подробнее
Номер записи: 3
16-05-2023 дата публикации

Repeated sports action counting method and device based on multi-scale transformation network

Номер: CN116129528A
Автор: GOU SHUIPING, LI RUIMIN, HUANG HU, HUANG LUGUANG
Принадлежит:

The invention relates to a repetitive sports action counting method and device based on a multi-scale transformation network, and the method comprises the steps: detecting skeleton key points in an original video to obtain an original skeleton sequence, carrying out the sampling of three scales of the original skeleton sequence, and obtaining three down-sampling skeleton sequences; carrying out bone joint point-by-bone joint point coding on the time periodic characteristics of each down-sampling bone sequence to obtain a time autocorrelation matrix of each bone joint; performing periodical selection from coarse to fine on the time autocorrelation matrix of each skeleton joint to obtain three weighted fusion period cubes; performing multi-scale periodic fusion on the three weighted fusion periodic cubes to obtain a multi-scale periodic cube; performing pulse diagram regression on the multi-scale periodic cube based on a multi-scale periodic transformation network to obtain a pulse diagram ...

Подробнее
Номер записи: 4
23-06-2023 дата публикации

Multi-spectral pedestrian detection method based on cross-modal feature enhancement and confidence fusion

Номер: CN116311364A
Автор: GOU SHUIPING, LIU XINYI, LI RUIMIN, XIANG JIAJUN, WANG XINLIN, LI SHENG
Принадлежит:

The invention discloses a multispectral pedestrian detection method based on cross-modal feature enhancement and confidence fusion. The multispectral pedestrian detection method mainly solves the problem that in the prior art, the missing detection rate and the false alarm rate of pedestrian detection are high under the single-modal condition. According to the implementation scheme, a visible light image and an infrared image are obtained to form a data set, and the data set is divided into a training set and a test set according to the proportion of 8: 2; constructing a double-flow pedestrian detection network formed by sequentially cascading an interactive common attention module, a pooling layer multi-scale adaptive fusion module and a confidence fusion module; using the training set to train the double-flow pedestrian detection network through small-batch gradient descent until the network loss converges; and inputting the test set into the trained network to obtain a multispectral ...

Подробнее
Номер записи: 5
23-05-2023 дата публикации

Hand posture and shape estimation method based on segmentation map guidance and regular constraint

Номер: CN116152199A
Автор: LI RUIMIN, LU HAOFAN, GOU SHUIPING, LI WEIBIN, ZHENG YANG, LIU HAOMIN
Принадлежит:

The invention discloses a hand posture and shape estimation method based on segmentation map guidance and regular constraint, and mainly solves the problem of low estimation result accuracy under the condition of hand object shielding in the prior art. According to the implementation scheme, a data set is divided into a training data set and a test data set; constructing a hand posture and shape estimation network based on segmentation map guidance and regular constraint, wherein the hand posture and shape estimation network is composed of a feature extraction trunk, a regular constraint pyramid module, a segmentation guidance module and a regression module; training the network by using a training data set; and inputting the test data set into the trained network to obtain a hand estimation result. According to the method, the segmentation guiding module is introduced to utilize the hand segmentation graphs with different visibility to assist in estimation under hand object shielding, ...

Подробнее
Номер записи: 6
25-07-2023 дата публикации

Intelligent pain management method and system based on facial expression recognition

Номер: CN116486461A
Автор: SU BINXIAO, WEI XINYI, GOU SHUIPING, MAO SHASHA, GU NAN, LI ANGZE, ZHANG HAOPENG
Принадлежит:

The invention discloses an intelligent pain management method based on facial expression recognition. The method comprises the following steps: S1, collecting facial videos under different pain degrees; s2, extracting face image data from the collected face video; s3, establishing an AI pain assessment model through deep learning, and assessing the degree of facial pain; in conclusion, by collecting the facial videos of the patient under different pain degrees and extracting the local facial expressions, key representations related to pain and the relevance between the key representations are conveniently captured, then data are processed and analyzed through the AI pain assessment model, the pain degrees are obtained, and a preliminary clinical medication strategy is provided according to the assessed pain degrees.

Подробнее
Номер записи: 7
27-06-2023 дата публикации

Maxillofacial region CT fracture detection and repair method based on geometric structure prior guidance

Номер: CN116342573A
Автор: TONG NUO, YANG HUIYING, GOU SHUIPING, LIU YUANLIN, HU XIAOYI, HOU LINGNAN
Принадлежит:

The invention discloses a maxillofacial region CT fracture detection and repair method based on geometric structure prior guidance, and mainly solves the problems that an existing method is not suitable for fracture bone repair and cannot well recover the occlusion function of a patient. According to the implementation scheme, CT data of a maxillofacial fracture patient is obtained, a fracture positioning data set and a skeleton contour data set are constructed, and a fracture detection model and a skeleton contour segmentation model are trained on the basis; constructing a simulated fracture defect skeleton contour training data set and a fracture skeleton test data set according to the trained model; constructing a maxillofacial region contour repair model based on geometric prior guidance self-attention learning, and training the model by using a simulated fracture defect skeleton contour training data set and using an adaptive learning rate optimization algorithm; and inputting the ...

Подробнее
Номер записи: 8
27-06-2023 дата публикации

Human skeleton action recognition method based on hierarchical space-time attention network

Номер: CN116343338A
Автор: LI RUIMIN, CHEN DONG, GOU SHUIPING, LI WEIBIN, ZHENG YANG, LU HAOFAN, HUANG HU
Принадлежит:

The invention discloses a human skeleton action recognition method based on a hierarchical space-time attention network, and mainly solves the problems that time features and space features are confused and multi-level action characterization is lacked in the prior art. According to the implementation scheme, a human skeleton sequence is obtained to form human skeleton data, and a training sample set and a test sample set are constructed; dividing articulation points in the training sample set and the test sample set into body parts, and dividing the human skeleton sequence into time slices; constructing a human skeleton action recognition model composed of a position coding module, a hierarchical space-time attention module and a classification module; training a human skeleton action recognition model by using the divided training sample set; and inputting the divided test sample set into the trained action recognition model to obtain a human skeleton action recognition result. The multi-level ...

Подробнее
Номер записи: 9
07-07-2023 дата публикации

Abdomen multi-organ increment segmentation method based on position guidance and consistency learning

Номер: CN116402800A
Автор: GOU SHUIPING, XU YINAN, LI MENGBIN, TONG NUO, HUANG LUGUANG, XU CHUNSHENG, LIAN BO, LIU BO, LIU WENBO
Принадлежит:

The invention discloses an abdominal multi-organ increment segmentation method based on position guidance and consistency learning, and mainly solves the problems of complex increment process and low increment performance in the prior art. According to the implementation scheme, firstly, a basic network is trained by using a consistency learning strategy to segment common abdominal organs such as liver, kidney, spleen, gallbladder and stomach for the first time; secondly, expanding the basic network to a position guiding increment network in a mode of newly adding a decoder, a position guiding module and a deep supervision module, and training the network to segment the liver, the kidney, the spleen, the gallbladder, the stomach and the pancreas for the second time; and expanding the position guidance incremental network to a position guidance and consistency learning incremental network, training the network to segment the liver, the kidney, the spleen, the gallbladder, the stomach, the ...

Подробнее
Номер записи: 10