22 results about "Training methods" patented technology

The invention relates to a speech interactive training system and a speech interactive training method. The system comprises a user selection module, a speech interactive training module, a user feedback module, a speech evaluation module and a result feedback module, wherein the user selection module is used for acquiring training contents selected by a user; the speech interactive training module is used for displaying the training contents to the user in a multimode guiding mode to guide the user to perform a speech training; the user feedback module is used for collecting a fed-back speech and a lip video corresponding to the speech; the speech evaluation module is used for receiving the speech fed back by the user and the lip video corresponding to the speech, and automatically evaluating the speech training of the user and giving an evaluation result; and the result feedback module is used for feeding the evaluation result back to the user so that the user can correct and adjust the speech training. The speech interactive training system is used for automatically evaluating the speech training of the user, giving the evaluation result and feeding the evaluation result back to the user, and then the user finds out the level of the personal speech training according to the evaluation result and corrects and adjusts the personal speech training to further improve the speech level, so the rehabitation training effect of a speech impediment is greatly enhanced.

The embodiment of the invention provides a training method for a mixed speech recognition model of mandarin and Sichuan. The method comprises the steps of determining features of mixed training audiodata, and determining the features as input data for training based on phoneme-based data alignment and Chinese character-based data alignment; inputting to N common intermediate layers, calculating afirst loss function by a first task layer, and calculating a second loss function by a second task layer; training N layers of first task layers based on the first loss function, training N layers ofsecond task layers based on the second loss function, performing multi-task training based on the trained first neural network parameters and the trained second neural network parameters, and training N layers of common intermediate layers. The embodiment of the invention further provides a training system for the mixed speech recognition model of mandarin and Sichuan. According to the embodimentof the invention, phonemes and Chinese characters are used as tasks of multi-task joint training, so that the recognition performance of mandarin and Sichuan talks is improved, and the resource occupation is reduced.

A device includes: a controller that converts a pressure value detected by a pressure detector into visual or aural lingual pressure information changed depending on a magnitude of the pressure value; and an information output unit, such as a buzzer and a color display LED, that outputs the lingual pressure information in response to a signal from the controller. The information output unit outputs lingual pressure information different depending on a pressing force applied on an elastic balloon by a tongue, allowing users such as patients and elderly people to continue a training without being bored.

The invention relates to the technical field of terminals, and provides a two-dimensional code display method based on a user behavior model, a user behavior model training method, electronic equipment and a storage medium. The two-dimensional code display method based on the user behavior model comprises the following steps: acquiring position information of a user and time information corresponding to the position information; inputting the position information and the time information into a pre-trained user behavior model, and determining a to-be-displayed target two-dimensional code; and controlling the target two-dimensional code to be displayed on the electronic equipment. According to the embodiment of the invention, the convenience of user operation can be improved, and the efficiency of two-dimensional code display is improved.

The invention discloses a model training and speech synthesis method and device, equipment and a medium, and relates to the technical field of artificial intelligence, deep learning and speech. The specific implementation scheme is that the model training method comprises the steps: obtaining a sample text in a training data set; determining label information corresponding to the sample text basedon an acoustic model trained by adopting an unsupervised training method in advance, wherein the label information comprises style information and/or role information; training a text classificationmodel based on the sample text and label information corresponding to the sample text, wherein the text classification model is used for outputting corresponding label information according to an input text. According to the embodiment of the invention, the technical effect of automatically determining the label information corresponding to the sample text is realized, the label labeling accuracyand efficiency are improved, and the training speed of the text training model is correspondingly improved.

The invention relates to a visual-angle-independent video three-dimensional human body posture recognition method. The method comprises the steps: 1, a virtual data generation stage: generating a two-dimensional/three-dimensional data tuple after synthesizing virtual camera parameters based on any human body posture data set containing three-dimensional marks; step 2, a model training stage: utilizing the generated two-dimensional/three-dimensional data tuples to train a modular neural network first module used for obtaining a model with camera visual angle generalization capability and a modular neural network second module used for obtaining a model capable of protecting inter-frame action continuity respectively; and 3, an unconstrained video reasoning stage: for any unconstrained acquired video, predicting by utilizing the multi-module deep neural network obtained by training in the step 2 to obtain a three-dimensional human body posture recognition result. Compared with the priorart, the method is based on a modular neural network combination training method, and the generalization ability of three-dimensional human body posture recognition is effectively improved.

The embodiment of the invention discloses a target vision model training method and device, equipment and a storage medium. The target vision model training method comprises the following steps: acquiring a first sample image and a background image; obtaining a target image according to the first sample image, and placing the target image in the background image to obtain a second sample image; inputting each sample image into a to-be-trained model, so that the to-be-trained model predicts a corresponding prediction category according to the sample image; and determining a first loss value according to the target category and the prediction category, adjusting model parameters of the to-be-trained model according to the first loss value until a preset training stopping condition is met, and taking the trained to-be-trained model as a target visual model. According to the embodiment of the invention, the problem of low model performance caused by few sample data, difficulty in sample data acquisition and the like can be solved.

The invention relates to a search processing method, a model training method, a device, a medium and equipment. The search processing method comprises the steps of receiving the target search information; determining a target search result according to the target search information, and determining target relevancy between the target search result and the target search information through a relevancy determination model; wherein the relevancy determination model is obtained by training in the following manner: according to historical operation behavior information implemented by a user for a plurality of historical search results, respectively determining relevancy between each historical search result and historical search information, and determining the relevancy between each historicalsearch result and the historical search information; wherein the relevancy between the historical search results with the same target text abstract information and the historical search information is the same; and training the model by taking the historical search information and the historical search result as the input of the model and taking the relevancy between the historical search resultand the historical search information as the target output of the model to obtain a relevancy determination model. In this way, the determined target relevancy is more accurate.

The embodiment of the invention provides a translation model training method and device and a translation method and device. The model training method comprises the following steps: respectively inputting a source language statement and a noisy source language statement in a parallel bilingual sentence pair into a translation model to obtain a first prediction target language statement and a second prediction target language statement, first prediction probability distribution and second prediction probability distribution of the translation model and/or first feature vectors and second feature vectors output by the hidden layers are obtained respectively; based on the first prediction target language statement and the target language statement in the parallel bilingual sentence pair, the second prediction target language statement and the target language statement corresponding to the noise-added source language statement, the first feature vector and the second feature vector and/or the first prediction probability distribution and the second prediction probability distribution; determining the current training loss of the translation model, and adjusting the parameters of the translation model. According to the embodiment of the invention, the robustness of the translation model can be improved, the training method is simple, and model training is stable.

The embodiment of the invention relates to the field of machine deep learning, and discloses a model training method, electronic equipment and a computer readable storage medium. The model training method comprises the following steps: performing migrating based on a preset pre-training model, and constructing a target model; obtaining a training set for training the target model; visualizing thetarget model, and determining a model structure of the target model; determining a to-be-frozen layer and a non-frozen layer of the target model according to the model structure, wherein the non-freezing layer is a network layer needing to be trained; and training the non-frozen layer according to the training set and the target model. According to the model training method provided by the embodiment of the invention, the feature information of the pre-training model can be highly utilized, the training effect of the target model is improved, and the risk of model overfitting is effectively avoided.

The embodiment of the invention is suitable for the technical field of computers, and provides a model training method and device, electronic equipment and a storage medium, wherein the model training method comprises the steps: shielding a part of image content of each first image in at least two first images, and obtaining at least two second images, wherein the at least two first images correspond to at least two image categories, and the similarity between the first images of different categories in the at least two first images is greater than a set value; training a first model based on the at least two first images and the at least two second images; and utilizing the first model obtained by training for identifying the image category of the input image.

The embodiment of the invention provides an object statistical model training method, and an object statistical method and device. The training method comprises: obtaining a training sample data set, wherein training samples in the training sample data set comprise training images and label data of the training images, the training images comprise multiple objects, and the objects comprise animals and/or plants; inputting the training image into the deep neural network model, and outputting a prediction density map; calculating a loss function according to the prediction density map and the label data to obtain a loss result; and iteratively adjusting network parameters of the deep neural network model according to the loss result, and generating a trained object statistical model.

The embodiment of the invention provides a motion key point video analysis model training method, and the method comprises the steps: providing a labeled video for a target key point analysis model, and obtaining a first result of motion key point analysis; and providing the labeled video to the trained teacher key point analysis model to obtain a second result of motion key point analysis, comparing the first result with a labeling result of labeled video data to obtain a first loss, and comparing the first result with a second result to obtain a second loss; synthesizing the first loss and the second loss according to a predetermined processing method to obtain a first synthesis loss; and training the target key point analysis model according to the first synthesis loss. According to the method, the trained teacher key point analysis model and the annotation result of the annotated video data are adopted to train the target key point analysis model. Therefore, the mobile client with the limited operational capability can obtain better analysis performance on the premise that the operational capability is not increased.

The invention provides a model training method based on market situation data sources. The model training method comprises the following steps of (1) data collection; (2) classified marking; and (3) training model acquisition. According to a sample classification method utilizing multiple data sources, multiple characteristic attributes of different data sources on the marked data and the unmarked data can be fully utilized in the implementing process, the method is suitable for the situation of multiple data sources, and the characteristic attributes of each data source can be ensured not to be acquired by other data sources in the whole training process.

The invention discloses a training method and training system for tissue lesion recognition based on an artificial neural network, and the method comprises the steps: preparing a training data set which comprises a plurality of examination images with lesions and annotation images which are related to the examination images and have lesion annotation results; performing feature extraction on the inspection image to obtain a feature map, and processing the inspection image based on an attention mechanism to obtain an attention heat map; using a first artificial neural network to classify the examination images, and combining the annotation images to obtain a first loss function, using a second artificial neural network module to classify the examination images based on the feature map and the attention heat map, and combining with the annotation result to obtain a second loss function, using a third artificial neural network to carry out disease-free judgment on the examination image to obtain a third loss function; by combining the three loss functions, the accuracy of tissue lesion recognition on tissue lesions can be effectively improved.

The invention discloses a semantic similarity task model training method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. Accordingto the specific implementation scheme, the method includes: when each piece of training data is adopted to train a semantic similarity task model, obtaining the score difference of positive and negative examples in the training data predicted by the semantic similarity task model; constructing an adaptive difference loss function of the semantic similarity task model based on the score differenceof the positive and negative examples in the training data obtained by a pre-trained semantic similarity scoring model and the score difference of the positive and negative examples in the training data predicted by the semantic similarity task model; and training the semantic similarity task model based on the adaptive difference loss function. According to the training method, the semantic similarity task model can learn signals of the semantic similarity scoring model more easily, and then the semantic similarity task model can effectively learn information of the semantic similarity scoring model.

This application discloses a training method and system for live work based on electric field visualization, which uses a combination of three-dimensional simulation technology, electric field simulation technology, virtual reality technology and motion capture technology to construct a three-dimensional simulation of live work in substations according to the actual substation operation scene Create a virtual scene and establish a simulation model of the electric field of the work scene, display the electric field of the work scene in the form of visual coloring, use wearable motion capture devices and force feedback devices to reflect the magnitude and position of the electric field intensity in the form of visual and tactile sensations, and let the work In the virtual scene, the personnel can see the changes of the electric field in the live work through real body movements, which is convenient for the operators to understand the relevant principles and knowledge of live work, improve the training effect, and solve the problem that the existing training and assessment methods for live work in substations cannot be intuitive Let the operators understand the relevant principles, and the technical problems of unsatisfactory training effects.