40 results about "Self supervision" patented technology

The invention provides a hospital information supervision platform and method. The hospital information supervision platform comprises a distributed block chain network constructed by adopting a plurality of data source parties and one or more data application parties as supervision nodes and further comprises a block chain key generation module, a distributed storage module and a sensing prompting module. The block chain key generation module adopts the block chain technology for enabling medical key information of the supervision nodes in the distributed block chain network to generate corresponding block chain keys through an encryption algorithm; the distributed storage module stores the block chain keys into all the supervision nodes of the distributed block chain network; the supervision nodes are communicated in the distributed block chain network; if medical key information of any supervision node is modified, the corresponding block chain key changes, and the other supervision nodes can sense the changes and give a prompt. The platform achieves multi-point self-supervision and intelligent prompting, the medical key information safety management cost is lowered, and the medical key information of a hospital can be prevented from being modified improperly.

The invention relates to a cross-modal joint hash retrieval method based on self-supervision. The method comprises the following steps: step 1, processing image modal data: carrying out feature extraction on the image modal data by adopting a deep convolutional neural network, carrying out Hash learning on the image data, and setting the number of nodes of the last full connection layer of the deep convolutional neural network as the length of a Hash code; step 2, processing the text modal data; using a word bag model for modeling text data, a two-layer full-connection neural network is established for feature extraction of text modal data, wherein the input of the neural network is a word vector represented by the word bag model, and the length of data of a first full-connection layer node is the same as that of data of a second full-connection layer node and a Hash code; step 3, for the neural network of category label processing, extracting semantic features from the label data by adopting a self-supervised training mode; and step 4, minimizing the distance between the features extracted from the image and the text network and the semantic features of the label network, so thatthe Hash model of the image and the text network can more fully learn the semantic features among different modals.

The invention puts forward an embedded attitude learning method capable of carrying out self supervision on the basis of a video time-space relationship. The average value of two auxiliary tasks is used for learning the time-space relationship in a video: one time sequence task learns whether two specific figure images are near or not on an aspect of time, and one spatial arrangement task learns ahuman body appearance model from space to enhance a capability of separating a gesture from a background. On the basis of the learning and mining repeated gesture of a course, training is carried outfrom a simple sample, then, the simple sample is iteratively expanded to a harder sample, and meanwhile, inactive video parts are eliminated. Time-space embedding successfully learns the representative characteristics of human gestures in a self-supervision way. By use of the method, a time-space relationship training video is used for carrying out gesture embedding self-supervision learning, manual annotation is not required, cost is lowered, gesture embedding can capture the visual characteristics of the human gestures, and human gesture estimation and retrieval efficiency can be improved.

The invention provides a college classroom teaching effect real-time monitoring and control system based on classroom multimedia equipment. The system includes a main controller, a sub-controller, a seat recognition device and a seat state detection device. The system only comprises a main controller, each seat needs to be correspondingly provided with a sub-controller, a seat recognition device and a seat state detection device; a face camera of the seat recognition device collects face images regularly, and a high-precision camera of the seat state detection device collects the images regularly. The method is expected to help teachers to easily and conveniently monitor the learning state of college students in class and carry out targeted intervention in time, protects the self-respect of the students to the maximum extent, enables the college students to accept the supervision, and helps the college students to improve the initiative and self-supervision of classroom learning.

The invention discloses an online education training system. The system comprises a cloud platform, a student learning terminal and a teacher management terminal. The cloud platform is used for transmitting network information data, student user information, teacher user information, learning material information and learning history information of the online education system are stored, and learning is intelligently arranged; the student learning terminal is used for login learning of students; the teacher management terminal is used for login teaching and management of teachers. Through theintelligent learning module, training, assessment, correction and review of students are automatically arranged, the learning progress and information of the students are monitored in real time, queryand viewing of teachers are facilitated, self-examination and self-supervision of the students are also facilitated, meanwhile, the students are arranged for communication and mutual assistance, andthe learning effect and the learning atmosphere are improved.

The invention relates to an image processing and image comparison model training method, device and system, a computer device and a readable storage medium. The method comprises the steps of obtaininga to-be-compared target image; inputting the target image into a pre-trained image comparison model, and outputting image features of the target image; wherein the image comparison model is obtainedthrough training based on a loss function including self-supervision loss, the self-supervision loss comprises loss between object classification data of different modal images of the same object in atraining sample, and the object classification data is obtained through classification processing based on image features; and comparing the image features of the target image with the image featuresof the base library images in the base library image group to obtain a comparison result. According to the method, the self-supervision loss is introduced for training, no extra marking cost is needed, and the cross-mode comparison accuracy is improved.

The invention teaches a personnel access control system and method for a copper wiring area. The method comprises the following steps of enabling pre-access personnel to enter a verification area, enabling a human body detection device to give out a triggering signal, performing pattern information acquisition on the pre-access personnel based on the triggering signal, judging whether the clothingof the pre-access personnel meets the specification or not according to pattern information, displaying the clothing judgment result to the pre-access personnel and opening or not opening an automatic door according to the judgment result. Compared with the prior art, the system and method can be used for solving the problem that when the personnel get in or out of the copper wiring area, only self-supervision is depended on and effective means do not exist to determine whether the clothing of the pre-access personnel meets the specification or not, copper pollution risk is effectively reduced, in addition, the verification of identity can be introduced as well, dual judgment on the qualification of personnel that get in and out of the copper wiring area is realized, management and control are better realized, meanwhile, the pattern information and identity information of the pre-access personnel of whom the clothing does not meet the specification or the verification of the identityis not passed can be recorded and stored as well, and the supervision and management are convenient.

A system and methods are provided in which an artificial intelligence inference module identifies targeted information in large-scale unlabeled data, wherein the artificial intelligence inference module autonomously learns hierarchical representations from large-scale unlabeled data and continually self-improves from self-labeled data points using a teacher model trained to detect known targets from combined inputs of a small hand labeled curated dataset prepared by a domain expert together with self-generated intermediate and global context features derived from the unlabeled dataset by unsupervised and self-supervised processes. The trained teacher model processes further unlabeled data to self-generate new weakly-supervised training samples that are self-refined and self-corrected, without human supervision, and then used as inputs to a noisy student model trained in a semi-supervised learning process on a combination of the teacher model training set and new weakly-supervised training samples. With each iteration, the noisy student model continually self-optimizes its learned parameters against a set of configurable validation criteria such that the learned parameters of the noisy student surpass and replace the learned parameter of the prior iteration teacher model, with these optimized learned parameters periodically used to update the artificial intelligence inference module.

The invention provides a platform convenient for doctor management and patient self-supervision, and aims to assist a patient to know a health condition and remind the patient to timely go back to a hospital for follow-up visit. The platform has main functions of patient archiving, examination index input, follow-up visit plan and rehabilitation plan pushing and reminding, related follow-up visit data analysis and the like, and provides online expert consultation and wardmate communication, thereby facilitating self-management and self-supervision of the patient. While follow-up visit and diagnosis and treatment process management is covered, data accumulated in an operation process is subjected to statistical analysis, so that big data value is comprehensively reflected and powerful data support and decision support are provided for clinic scientific and research of nephrology departments. The follow-up visit is taken as a center; for the patient with the chronic kidney disease, the follow-up visit is performed in the way; the problem that a kidney disease assistant in China at present takes no count of the follow-up visit is solved; the burden of a doctor and the frequency of going to the hospital for the patient are reduced; and a few unnecessary expenses are saved.

The invention discloses a system and a method for monitoring automobile scratch based on cloud computing. The system comprises a sensor module, a control module, a camera module, a display module, a local storage module, and a cloud server. The control module is connected with the sensor module, the camera module, the display module, and the local storage module, respectively. The local storage module adopts a storage device built in the automobile central control system, and is connected with a cloud server network through the automobile central control system. The cloud server stores the information transmitted by each automobile, and opens the reading authority to the automobile owner, the traffic control department, and the insurance department. The system and the method of the invention can solve the defect of the automobile self-supervision system, and overcomes the disadvantage that the service monitoring position is fixed. In addition, the system and the method of the inventioncan avoid or reduce the loss caused by scratch of the automobile, and upload the scratch video of the automobile to a cloud computing data processing center.

The present invention provides a supervision and control method for UAV (unmanned aerial vehicle) threat avoidance by a manned vehicle, relating to the field of UAV. According to the method, a supervisory control mode, an environmental factor and threat avoidance action are defined and corresponding values are defined, the inference and judgment of optional threat avoidance actions is carried out, then a final threat avoidance action is judged according to a judgment result, and the threat avoidance is completed. For a known threat, the UAV fully plays independence ability to independently complete threat avoidance or completes operation under the guidance of the manned vehicle. For an unknown threat, the UAV completes avoidance under the guidance of the manned vehicle, and the autonomous execution ability of the UAV and the analysis and judgment ability of a manned vehicle operator are fully played. For an extreme case such as the interruption of communication, the UAV ensures its own safety through adjusting the supervisory control mode. A variable self-supervision and control mode is combined with the characteristics of UAV and the manned vehicle, and the method has good action ability in emergency for different threat types and environmental conditions.

The invention relates to a medical platform based on a block chain. A block chain key generation module generates corresponding block chain keys from medical key information of supervision nodes in adistributed block chain grid through an encryption algorithm by adopting a block chain technology; a distributed storage module stores the block chain key into all supervision nodes of a distributed block chain grid, wherein the supervision nodes communicate with one another in the distributed block chain grid; if the medical key information of any supervision node is modified, the block chain keygeneration module generates a corresponding block chain key change, and other supervision nodes perceive and prompt through the perception prompt module; and after the perception prompting module carries out prompting, the coverage module covers the block chain key with the change of the supervision node by using the previously stored block chain key when the set duration is exceeded. Compared with the prior art, self-supervision, intelligent prompt and data recovery can be achieved.

The invention belongs to the technical field of computer application, and discloses a data processing system and method for chronic disease detection based on a computer prediction model. The method comprises the steps: obtaining various types of data of a user, and transmitting the collected data through an Internet system; classifying and summarizing the received various data of the user, and transmitting the data through the Internet system; analyzing and predicting the data according to the computer prediction model, and transmitting a result to an output module. On the basis of the computer prediction model, the system completes the summarization, analysis, arrangement and prediction of various data of the user, so that the user can conveniently perform more effective self-supervisionmanagement and professionals can conveniently give more accurate health promotion suggestions by integrating the data, thereby effectively preventing hypertension chronic diseases; the prediction efficiency can be improved according to the hypertension chronic disease prediction value.

The invention provides a method and device of a taxi industry driver order escape behavior recognition system. The device comprises a taxi and a taxi enterprise management platform, wherein the taxi is internally provided with a bridge, a driver's seat and back seats, the bridge, the driver's seat and the back seats are sequentially arranged from front to back, the bridge is equipped with a vehicle taximeter and vehicle-mounted terminal equipment, the vehicle-mounted terminal equipment includes a multimedia module and an anti-order escape behavior recognition APP, the anti-order escape behavior recognition APP includes a vehicle speed monitoring module, a face recognition algorithm resource library, a taximeter state detection module, a taxi driver order escape recognition algorithm logicmodule and a fault-tolerant module, and the top of the taxi is equipped with a GPS antenna. The method and device of the taxi industry driver order escape behavior recognition system provides a brandnew supervision weapon for taxi enterprises and government departments, and greatly facilitates the regulation performed by the government on the taxi industry and self-supervision of the taxi enterprises.

The invention provides a self-supervision-based visual representation learning method. Main contents of the method include: self-supervision, visual representation, image construction and transformation learning in an image. A process of the method includes: adopting a self-supervision method to learn invariance of which expression includes differences between instances and differences inside theinstances; building a graph, in which nodes indicate image patches, describing similarities between the image patches, and defining two types of edges associating the image patches mutually in the graph, wherein constructing the graph with the instances and the internal edges includes scaling moving objects and is through the edges between the clustered instances and the edges inside the tracked instances; and training deep neural networks to generate similar visual representation. The visual feature learning method of the invention carries out self-supervision, can automatically acquire annotation labels, and greatly saves manpower and material resources; and reduces, at the same time, impacts brought by viewpoints, postures, deformation, illumination and the like, and increases an accuracy rate.

The invention provides a neural network training method and device, and the method comprises the steps: obtaining a plurality of feature maps outputted by a plurality of intermediate layers of a to-be-trained neural network; performing feature extraction on the plurality of feature maps through a feature extraction network to obtain first feature output of each intermediate layer; according to theplurality of first feature outputs and second feature outputs output by the to-be-trained neural network, calculating to obtain a first loss; based on the first loss, parameters of the plurality of intermediate layers are adjusted. Through a self-distillation mode, each intermediate layer of the neural network model and result features extracted in each iteration are output to serve as self supervision signals to be fully utilized, a result can be converged more quickly, training is completed, and time and resources are saved.

A computer-implemented method to improve scale consistency and / or scale awareness in a model of self-supervised depth and ego-motion prediction neural networks processing a video stream of monocular images, wherein complementary GPS coordinates synchronized with the images are used to calculate a GPS to scale loss to enforce the scale-consistency and / or -awareness on the monocular self-supervised ego-motion and depth estimation. A relative weight assigned to the GPS to scale loss exponentially increases as training progresses. The depth and ego-motion prediction neural networks are trained using an appearance-based photometric loss between real and synthesized target images, as well as a smoothness loss on the depth predictions.

The invention discloses an intelligent video image retrieval method based on a neural network self-temperature fault and a knowledge conduction mechanism, which improves the retrieval precision of a small model while ensuring the real-time performance of the model, and achieves balance between the precision and the efficiency as far as possible. A gamma correction module is arranged, and through local adjustment of the image, illumination non-uniformity robustness is achieved, detail discernibility is improved, high-frequency noise is avoided, and universality is high; a self-temperature fault mechanism is established, local self-supervision, continuous reflection and learning parameter adjustment of the neural network are allowed, deep semantic information of the image is fully learned, rapid convergence of the neural network is achieved, and retrieval precision is improved; a knowledge conduction mechanism is adopted, the model precision is improved, the model time delay is reduced, network parameters are compressed, and finally a student model with high performance and high precision is obtained; and taking shallow feature knowledge as a learning target through a conduction mechanism, reconstructing deep features by adopting a VAE variational self-encoding model so as to generate a learning result, and measuring the learning result and the target so as to complete a learning task.

The invention discloses an information management system suitable for an enterprise. The information management system comprises a server, a terminal management module, a decision approval module, a data management module, a private reporting module, an announcement display module and a user permission classification module. The information management system is characterized by establishing a complete information sharing system, toolizing a management system through a software technology and a network technology, and establishing a smooth information communication system, an effective cooperative execution system and an accurate decision support system in an enterprise to improve the management and office capacity in an organization; by establishing a private reporting module, facilitatinga highest leader end to obtain direct feedback information, thus being able to know about internal defects of the enterprise, being convenient for optimization and adjustment of an enterprise system,decisions and the like; directly publicizing related information through the announcement display module, so that self-supervision and self-progress in employees can be established; and by establishing a data management module, facilitating data transfer among enterprises, and improving the information transmission efficiency.

The invention provides a control system of an explosion-proof forklift, and further provides a self-supervision method of the explosion-proof forklift. The explosion-proof forklift control system comprises a sensor, a system controller and an on-board upper computer. The system controller comprises an information processing module, a memory module, a battery module, a data acquisition module, a communication module and a position recorder device. The data acquisition module, the memory module, the communication module and the position recorder device are in signal connection with the information processing module. The battery module is electrically connected with the information processing module, the memory module, the data acquisition module, the communication module, the position recorder device, a speed sensor, a temperature sensor and an electric spark sensor. According to the self-supervision method of the explosion-proof forklift, the sensors detect the state of the explosion-proof forklift and send the state to the system controller, and meanwhile, the system controller controls the explosion-proof forklift to automatically take corresponding treatment measures based on thestate of the explosion-proof forklift, so that accidental loss is reduced, and the probability of occurrence of safety accidents is reduced.

Methods and systems disclosed herein relate generally to systems and methods for generating visual relationship graphs that identify relationships between objects depicted in an image. A vision-language application uses transformer encoders to generate a graph structure, in which the graph structure represents a dependency between a first region and a second region of an image. The dependency indicates that a contextual representation of the first region was derived, at least in part, by processing the second region. The contextual representation identifies a predicted identity of an image object depicted in the first region. The predicted identity is determined at least in part by identifying a relationship between the first region and other data objects associated with various modalities.

The invention provides a control system of an explosion-proof forklift, and further provides a self-supervision method of the explosion-proof forklift. The explosion-proof forklift control system comprises a sensor, a system controller and an on-board upper computer. The system controller comprises an information processing module, a memory module, a battery module, a data acquisition module, a communication module and a position recorder device. The data acquisition module, the memory module, the communication module and the position recorder device are in signal connection with the information processing module. The battery module is electrically connected with the information processing module, the memory module, the data acquisition module, the communication module, the position recorder device, a speed sensor, a temperature sensor and an electric spark sensor. According to the self-supervision method of the explosion-proof forklift, the sensors detect the state of the explosion-proof forklift and send the state to the system controller, and meanwhile, the system controller controls the explosion-proof forklift to automatically take corresponding treatment measures based on thestate of the explosion-proof forklift, so that accidental loss is reduced, and the probability of occurrence of safety accidents is reduced.

The invention discloses a student extracurricular reading supervision method and system based on an electronic book. The method comprises steps of reading behavior data generated by a user in the reading process being acquired; wherein the reading behavior data comprises one or more of a single reading total duration, a reading total word number, a time interval of adjacent page turning behaviorsand annotations during reading; acquiring physical data of user equipment; wherein the physical data comprises real-time image data, equipment gravity sensor data and position sensor data; detecting the reading behavior data and the physical data of the user, judging whether abnormal behaviors exist or not, and generating abnormal reading data and an abnormal reading conclusion. The method is advantaged in that the comparison between the adjacent page turning data read this time and the last non-abnormal reading data is more comparable, so a manager can better manage the user conveniently, andmeanwhile, the self-supervision is facilitated.

A system and methods are provided in which an artificial intelligence inference module identifies targeted information in large-scale unlabeled data, wherein the artificial intelligence inference module autonomously learns hierarchical representations from large-scale unlabeled data and continually self-improves from self-labeled data points using a teacher model trained to detect known targets from combined inputs of a small hand labeled curated dataset prepared by a domain expert together with self-generated intermediate and global context features derived from the unlabeled dataset by unsupervised and self-supervised processes. The trained teacher model processes further unlabeled data to self-generate new weakly-supervised training samples that are self-refined and self-corrected, without human supervision, and then used as inputs to a noisy student model trained in a semi-supervised learning process on a combination of the teacher model training set and new weakly-supervised training samples. With each iteration, the noisy student model continually self-optimizes its learned parameters against a set of configurable validation criteria such that the learned parameters of the noisy student surpass and replace the learned parameter of the prior iteration teacher model, with these optimized learned parameters periodically used to update the artificial intelligence inference module.

The invention discloses a method for extracting relations between entities from a text based on self-supervision and clustering technologies, and the method comprises the steps: carrying out the relation extraction in a label-free data set based on the idea of self-supervision. A context encoding module, a clustering module and a classification module are included. The context encoding module is used for encoding the data and then performing relationship extraction and analysis; the clustering module is divided into two parts of adaptive clustering of entities and adaptive clustering of entitycategories, wherein the adaptive clustering of the entities is used for directly clustering the entity codes, and the obtained clustering result is an entity category, wherein the self-adaptive clustering of the entity categories is to cluster the entity categories, and the obtained result is the relationship between the entity categories; and the classification module is used for linking all themodules for joint training.

The invention discloses an exercising process self-supervision platform based on the internet and a using method. The platform comprises two portions: a client module I and a server side module II; the client module I takes charge of member registration, member logging-in, post-exercising self-supervision result input and self-supervision result analysis display; the server side module II mainly takes charge of member registration, legal member verification, member post-exercising self-supervision result storage and analysis on a self-supervision result after a period of exercising. According to the invention, in a case of not requiring intervention of a professional doctor, self-supervision can be carried out according to an experience result obtained after a sporter exercises, and a supervision result can be analyzed after a period of exercising so as to determine whether a current exercise plan or exercise load is suitable; the exercising process self-supervision platform is helpful for preventing over fatigue of the sporter; the basis is provided for the sporter to scientifically take part in physical learning and exercise; reference is also provided for physical examination of a doctor.

The invention relates to the technical field of airport baggage handling, in particular to a baggage confirmation method based on an electronic baggage tag. Through the method provided by the invention, the airport carries out sectional management on luggage consignment, each section of node is confirmed, and meanwhile, the responsible person in each section of luggage transfer process can be conveniently confirmed. In the luggage transferring process, if the luggage is confirmed in a certain link and is not found, it can be known that the luggage is wrong in the operation process of the link,and then the luggage cannot be found. Through a definite responsible person system of each link, a system of layer-by-layer control and layer-by-layer responsible is achieved, self-supervision of workers in the baggage transferring process is facilitated, responsibility awareness of the workers is enhanced, and the baggage transferring efficiency is improved. Meanwhile, the luggage transfer efficiency is further improved, and more importantly, the luggage loss can be greatly reduced.

The products relates to a banking service evaluation system, which is innovation and upgrading of the traditional window evaluation machine, wherein customer evaluation is longer limited by the window part, the evaluation is expanded to evaluation on environment and safety of the banking hall and service efficiency, and the self-monitoring and self-management mechanism of the banking system can be improved with the appearance arrangement of the bank numbering machine so as to perfect banking service. The banking service evaluation machine comprises: (1) a housing, (2) a power supply circuit, (3) an embedded minimum system, (4) an auxiliary circuit, (5) a liquid crystal screen interface circuit, (6) a touch screen interface circuit, and (7) a network communications circuit. The product can be provided for well feeding back omnidirectional evaluation of customers on banking service.

The invention provides a real-time monitoring and control system for university classroom teaching effect based on classroom multimedia equipment, including a main controller, a sub-controller, a seat identification device and a seat state detection device, including only one main controller, each seat needs Correspondingly configure a sub-controller, a seat recognition device and a seat state detection device, the face camera of the seat recognition device regularly collects face images, and the high-precision camera of the seat state detection device periodically collects images. collection. The invention is expected to help teachers easily and conveniently monitor the learning status of college students in the classroom, and carry out targeted intervention in a timely manner, which protects the self-esteem of students to the greatest extent, makes college students more willing to accept such supervision, and helps college students improve classroom learning. Proactivity and self-monitoring.

The present disclosure relates to devices and methods for determining new virtual evidence to use with a deep probabilistic logic module. The devices and methods may receive output from a deep probabilistic logic module in response to running an initial set of virtual evidence through the deep probabilistic logic module. The devices and methods may use the output to automatically propose at least one factor as new virtual evidence for use with the deep probabilistic logic module. The devices and methods may add the new virtual evidence to the deep probabilistic logic module.