522 results about "Latent vector" patented technology

The present invention provides a system and method for causality-augmented generative intelligence capable of autonomously discovering non-obvious actionable insights from heterogeneous and multimodal data sources. The system integrates a data ingestion unit for semantic and temporal harmonization of structured and unstructured datasets, a causal inference processor for constructing a dynamically evolving directed causal knowledge representation using perturbation-based validation, a latent representation processor that combines multimodal semantic embeddings with causal parameters to generate fused latent vectors, and a generative insight processor utilizing causally constrained generative reasoning to synthesize hypotheses anchored to verified cause-effect dependencies. A validation processor performs counterfactual assessment and observational verification to ensure retention of only those insights that remain consistent with causal ground truth.

Provided is a real scene image editing method based on hierarchically classified text guidance, including: firstly selecting a hierarchical multi-label text classification model and hierarchically classify an input style description text; obtaining a latent vector of an indoor scene image and dividing the latent vector; training latent space residual mappers which are divided into four groups for generating details of a layout, an object, an attribute, and a color in the scene image, and selectively training a mapping model with a secondary word obtained by a text classification model; inputting a tertiary word obtained by the text classification model to a contrastive language-image pre-training (CLIP) network and controlling training of the mapping network by utilizing a CLIP loss; hierarchically inputting the latent vector to the mapping network to obtain a bias vector, summing the bias vector with an original vector for inputting to the StyleGAN to obtain an edited image.

The invention provides a robot motion control model training method and device based on deep reinforcement learning, and relates to the technical field of sensors and robots. The method comprises the following steps: performing deep reinforcement learning-based training on a strategy network for outputting an action strategy for controlling the robot to move by utilizing a teacher-student model framework; wherein the training process of the strategy network further comprises the steps of encoding historical linear velocity information of the robot through a linear velocity encoder to generate a first potential vector, and inputting the first potential vector into the strategy network for auxiliary training. According to the method, the motion trend, the state change track and the surrounding terrain structure of the robot are comprehensively considered during action decision making, and strategy output is more stably and accurately controlled.

Systems and methods for generating tile-able patterns from text include obtaining a text prompt and generating, by a generation prior model, a latent vector based on the text prompt, where the generation prior model is trained to output vectors within a distribution of tile-able patterns. An image generation model then generates an output image based on the latent vector. The output image comprises a tile-able pattern including an element from the text prompt.

The invention discloses a long video role scene decoupling generation method and system based on multi-modal diffusion, and relates to the technical field of image processing, and the method comprises the steps: S1, synthesizing the advanced features of a role and a scene through a SigLIP encoder and a DINOv2 encoder; s2, performing cross-modal feature fusion on the advanced features to obtain joint features, and compressing the joint features to obtain compact vectors; s3, generating text features according to the text prompt; s4, potential codes are generated from an input video through a causal 3D convolution encoder, the potential codes pass through a linear projection matrix and then are spliced with a memory state for dimension reduction, and a segmented potential vector sequence is obtained; s5, performing decoupling perception generation on the segmented potential vector sequence through an improved 3D-UNet, and performing deconvolution up-sampling reconstruction after deterministic sampling to obtain an RGB video segmented sequence; according to the method, the key problems of rough dynamic control, limited generation length and over-high resource consumption in long video generation are solved, and the quality and efficiency of the generated video are remarkably improved.

The present invention provides a video generation method, comprising: encoding an image comprising a reference figure to obtain a feature representation of the reference figure; encoding a video comprising a reference action to obtain a feature representation of the reference action; injecting the feature representation of the reference figure and the feature representation of the reference action into a latent diffusion network in a video latent diffusion model by means of at least one timing hold module in a trained video latent diffusion model, wherein the latent diffusion network comprises a plurality of cross-attention and denoising units, one timing hold module is connected in series between at least two cross-attention and denoising units, and each timing hold module comprises a time converter and a space converter; using the latent diffusion network to denoise an inputted noise sequence to obtain a denoised latent vector sequence; and decoding the latent vector sequence to obtain an action video of a target object. The present invention further provides a video generation apparatus, an electronic device, a storage medium, and a program product.

The embodiment of the invention provides a character animation generation method and device based on voice driving, equipment and a medium, and belongs to the technical field of financial science and technology. The method comprises the following steps: acquiring an audio signal input by a user, a target human portrait picture and a target scene reference video; encoding the audio signal to obtain an audio code; generating a three-dimensional face model motion coefficient according to the portrait picture of the target person and the audio code; extracting blink information and head posture information from the target scene reference video; correcting the motion coefficient of the three-dimensional face model according to the blinking information and the head posture information to obtain a corrected motion coefficient of the three-dimensional face model; and taking a three-dimensional deformable head model as a potential vector, mapping the three-dimensional deformable head model to the corrected three-dimensional face model motion coefficient frame by frame, converting the three-dimensional face model motion coefficient into a video frame, and generating a head speech video of the target person. According to the embodiment of the invention, the quality and reality of character animation generation driven by voice can be improved.

Systems, methods, and instrumentalities are disclosed for video encoding and / or video decoding using artificial neural networks (e.g., convolutional neural networks or recurrent neural networks), attention, and / or attention with spatial attributes. For example, an apparatus may be configured to perform one or more of the following: obtaining a context block, a current block, and a latent vector associated with the context block; performing at least one convolution on the context block, the reference block, and the latent vector; generating motion flow data associated with the current block based on the at least one convolution; or generating a bitstream the comprises an indication of the motion flow data. The motion flow data may be quantized. The generated bitstream may comprise an indication of the quantized motion flow data.

The invention provides a spatial intelligent three-dimensional modeling method based on a two-dimensional structure design sketch image, and the method comprises the steps: carrying out the preprocessing of an input two-dimensional structure design sketch image, and obtaining a preprocessed design sketch image, extracting two-dimensional structure design features from the preprocessed design sketch image, and encoding the two-dimensional structure design features to generate structured tensor representation; deducing space components, function partitions and constraint logic based on structured tensor representation to construct a sketch semantic graph, nodes of the sketch semantic graph representing physical structure units, and edges of the sketch semantic graph representing connection relations and stress constraints among the physical structure units; determining a node embedding vector corresponding to each node in the sketch semantic graph so as to perform three-dimensional space coding on the node embedding vector to obtain a node potential vector, and generating a geometric reasoning sequence during three-dimensional structure geometric reasoning according to a mechanical logic definition on the basis of edges in the sketch semantic graph, and performing three-dimensional structure geometric reasoning based on the node potential vectors and the geometric reasoning sequence to generate a three-dimensional structure model.

The invention relates to a method for generating an image watermark based on potential space multi-scale feature modulation. The method mainly comprises a watermark residual embedding and extracting method based on potential space multi-scale feature modulation, a watermark intensity regulation and control strategy based on time sequence gating and a robust watermark end-to-end training framework based on potential space. According to the method, watermark information is fused by using a multi-scale structure in a potential space of a diffusion model, meanwhile, a time sequence gating strategy dynamically determines the embedding time step of a watermark by calculating the convergence degree of a potential vector, and the image quality and the watermark robustness are balanced. Sigmoid weighted noise prediction loss is introduced in the model training process, and denoising control over low time steps is enhanced. The method can be widely applied to scenes such as image generation, copyright identification and content traceability, and has good practicability and popularization prospects.

According to the single-image geometric texture joint control image generation method and the terminal, an original image and text prompt information transmitted by a user are received, and a human face sub-image, a key point image and a front face image are generated based on the original image; performing feature extraction and processing on the human face sub-images to generate human face identity implicit vectors; generating a hidden space control vector based on the key point graph and the frontal face image; generating a text embedding feature based on the text prompt information; injecting the human face identity implicit vector, the text embedding feature and the implicit space control vector after splicing the noise vector into a diffusion converter model by adopting a hierarchical selective feature injection mode, and generating an identity holding image by combining a decoder; the diffusion converter model is obtained through training by adopting a low-rank adaptive training strategy; and controllable generation of the identity maintaining image with single image driving and multi-pose styles is realized.

The invention provides a robot motion control model training method, device and equipment based on deep reinforcement learning, a robot and a medium, and relates to the technical field of robots. The method comprises the following steps: acquiring a first potential vector obtained after a student encoder encodes robot body observation data, and a second potential vector obtained after a teacher encoder encodes privilege observation data; based on the current training step number and a preset probability function, calculating a sampling probability for controlling a fusion proportion of the first potential vector and the second potential vector; fusing the first potential vector and the second potential vector based on the sampling probability to generate a third potential vector, and inputting the third potential vector into a strategy network; and updating the parameters of the policy network based on the value estimation of the current state output by the value network and the action policy output by the policy network. According to the method, updating oscillation caused by sudden change of input distribution in the training process of the strategy network can be avoided, the training efficiency is improved, and the training cost is reduced.

A method for detecting an anomaly, which is performed by a computing device including one or more processors according to some embodiments of the present disclosure may include: collecting a CAN message generated in a controller area network (CAN); generating first traffic time interval data and first payload data based on the CAN message; obtaining a first latent vector generated by inputting the first traffic time interval data into a first neural network model; obtaining a second latent vector generated by inputting the first payload data into a second neural network model; generating merged data in which the first latent vector and the second latent vector are merged; and determining whether the CAN message is anomalous by inputting the merged data into a third neural network model.

The invention discloses a CO2 geological sequestration inversion simulation method based on a deep learning parameterization strategy, and the method comprises the steps: carrying out the parameterization dimension reduction representation of a complex permeability field of a CO2 geological sequestration storage layer through a deep learning method, and representing the distribution characteristics of a heterogeneous permeability field through a group of latent vectors in low-dimensional normal distribution, and the number of to-be-estimated parameters in CO2 geological sequestration numerical simulation is reduced. According to the method, a parameterization method based on deep learning and a multi-data assimilation iteration set smoothing algorithm are combined, the heterogeneous permeability field of a complex reservoir stratum can be efficiently estimated only by cyclically updating the low-dimensional latent vector during inversion, and then the simulation and prediction precision of a numerical model on the CO2 geological storage process is improved. According to the method, by effectively estimating the complex permeability distribution of the CO2 geological storage layer, the simulation description precision of the CO2 migration and transformation process is improved, and technical support can be provided for optimized operation and risk management of a CO2 geological storage project.

A method that includes obtaining, from one or more stations, embedding vectors that embed features of the stations, obtaining measurement vectors and associated measurement names, generating a text array of the measurement names utilizing a language model, concatenating the text array and the measurement vector at one or more cross-attention modules configured to encode one or more measurement arrays to one or more latent embedding vectors, generating one or more latent embedding vectors associated with the measurement vector and corresponding measurement names via the cross-attention module and a fixed-size station embedding vector, outputting the latent embeddings; generating a query vector; generating key vectors value vectors utilizing a latent embedding vector; decoding the latent vectors utilizing the key vector and value vector; utilizing the cross attention modules and query vectors, decoding the latent embedding vectors; and output a predication.

The invention provides a face image restoration method based on generative adversarial network inverse mapping, and belongs to the technical field of image restoration and computer vision. According to the method, an encoder-decoder architecture is adopted, an encoder is a ResNet combined with a channel attention mechanism and can encode a to-be-restored face image (containing block-shaped shielding) to a potential space of a StyleGAN, and a W + vector of 18 * 512 dimensions is obtained; and the decoder is a pre-trained StyleGAN generator, and can convert the potential vector into a complete face image to realize restoration. In the training process, through a weighted combination constraint model of L2 loss, perception loss and face identity loss, it is ensured that a restoration result is consistent with an original image in pixel, feature and identity levels. Meanwhile, by means of the decoupling characteristic of the StyleGAN potential space, the repaired face features (such as smile and age) can be edited. According to the method, complex preprocessing is not needed, large-area missing face images can be efficiently repaired, identity consistency is kept, and the method is suitable for monitoring image enhancement, old photo repair and other scenes.

The invention discloses an adversarial patch generation method based on a sensitivity graph. The method comprises the following steps: firstly, extracting potential vectors from standard normal distribution, inputting the potential vectors into a pre-training generator to generate an initial adversarial patch, carrying out gradient descent optimization, combining with a specific loss function, and superposing the initial adversarial patch to a target image through various transformations to verify robustness and effectiveness in different environments; and finally, calculating imperceptibility loss and calculating adversarial detection loss and total variation loss by calculating a sensitivity graph and creating a mask matrix to cut a patch, so as to restrain generation of an adversarial patch sample. The adversarial patch generated by the method has high effectiveness and concealment, has strong robustness in different environments, can effectively deceive the target deep learning model, provides powerful support for safety assessment of the deep learning model, and is beneficial to improving the safety protection capability of a deep learning system.

The invention discloses an adversarial texture generation method and system applied to a depth estimation model. The method comprises the following steps: generating an initial adversarial texture pattern based on a potential vector driving pattern generator; performing up-sampling processing on the initial confrontation texture pattern to obtain an up-sampling texture pattern; mapping the up-sampling texture pattern to a specified area of the object model, and performing image rendering on the mapped object model M by using a rendering engine to generate an object image; performing background synthesis on the object image to generate a complete scene attack image; inputting the complete scene attack image into the depth estimation model to obtain the predicted depth of the image at the current view angle; and designing a loss function for optimization based on the predicted depth and the reference depth. According to the scheme, based on the potential vector driven generator, texture mapping and background synthesis processing, the method is not limited to a two-dimensional image task any more, the training efficiency is improved, and the diversity and stability of generated textures are improved.

The embodiment of the invention discloses an image editing and generating method and device, electronic equipment and a storage medium, and relates to the technical field of image generation, and the method comprises the steps: obtaining a to-be-processed image and a target instruction, constructing an image description of the to-be-processed image according to a scene in the to-be-processed image, converting the image description into a text mark, and storing the text mark in the target instruction; the method comprises the steps of obtaining a to-be-processed image, embedding a text mark into a high-dimensional space through an embedding model to obtain text embedding, obtaining an element vector of the to-be-processed image according to the to-be-processed image and the text embedding through a diffusion model, designing an editing operator according to a target instruction, updating the element vector by using the editing operator, and generating a result image according to the updated element vector through the diffusion model. According to the method, efficient, accurate and flexible editing of the image is realized by combining the diffusion model and the editing operator, and the problems that the expression of the model on a specific target category cannot be improved, the problem of long tail distribution cannot be solved and data similar to a real scene cannot be generated in the prior art are solved.

The invention provides a method and device for generating an image based on multi-scale feature fusion, and the method comprises the steps: obtaining text features and target noise, the text features being used for providing content guidance for generating the image; the target noise is processed through a trained target encoder, a potential vector fused with multi-scale features is generated, the multi-scale features are used for indicating details and overall structure information of different levels in a generated image, and the potential vector is used for indicating image features of the generated image; the text features are fused into the potential vectors through a trained target fusion model, fusion features are generated, and the fusion features are used for achieving generation of a text guide image; and decoding the fusion features into a final image through a trained target decoder. The quality of the generated image can be improved.

The invention discloses a confrontation trajectory prediction method, a readable storage medium, electronic equipment and a vehicle, and belongs to the field of automatic driving. The prediction method comprises the following steps: constructing a trajectory prediction model; inputting a historical track, a real track and environment context information; track features are obtained through feature extraction and a space-time attention mechanism; constructing a confrontation model of the explicit space; mapping the track features to a submerged space; disturbance is carried out on the submerged vectors, and the submerged vectors are regularized to specific distribution; and mapping the latent vector to the explicit space, and generating a confrontation trajectory of the target driving vehicle. The readable storage medium stores a program of the method. A program including the method is stored in the electronic equipment. The vehicle comprises the intelligent system for executing the method. According to the method, the explicit space and the implicit space are emphasized for multi-space collaborative optimization, the pertinence and the real concealment of attacks are remarkably improved, the prediction precision is improved, and then the target driving vehicle track can resist surrounding vehicles and cope with complex vehicle conditions.

The invention discloses a depth time sequence prediction method and system based on an event gating mechanism, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining time sequence data and corresponding event sequence data; mapping the time sequence data and the corresponding event sequence data to a low-dimensional potential vector space to obtain a time sequence feature representation and an event sequence feature representation; according to the time step sequence, inputting the time sequence feature representation and the event sequence feature representation into an improved recurrent neural network for processing to obtain a hidden state sequence; performing attention calculation based on the hidden state sequence to obtain final vector representation; and obtaining a prediction result according to the final vector representation. According to the invention, the prediction capability of the time sequence containing the non-periodic event interference can be improved.

A latent transformer architecture with latent attention mechanisms and expert processing systems for federated deep learning is disclosed. The system operates entirely within latent space, eliminating traditional embedding and positional encoding layers while maintaining full attention capabilities. Input data is compressed into latent vectors via variational autoencoder encoding, then processed by a latent attention module that computes query, key, and value matrices directly from latent representations. The architecture incorporates expert processing systems including gated latent expert networks for sparse computation and latent mixture of experts for collaborative processing. In the gated approach, a routing network selectively activates specialized expert modules based on latent vector characteristics. The mixture approach enables all experts to contribute through weighted combination, facilitating distributed computation and enhanced model expressiveness.

An information processing apparatus 4 includes: a transformation unit 423 configured to transform a plurality of tuples of feature vector data XV in a presentation space into a plurality of tuples of latent vector data ZV in a latent space; an extraction unit 423 configured to extract, based on the plurality of tuples of latent vector data, feature vector data having a shorter distance from query data DQ in the latent space than the other feature vector data, as neighbor data DNi from among the plurality of tuples of feature vector data; a generation unit 424 configured to generate, based on the neighbor data, a local model LM that outputs an estimation value dpi of a latent distance di when difference information Vi is inputted, the latent distance being a distance between the query data and the neighbor data in the latent space, the difference information being related to a difference, for each element of the features, between the query data and the neighbor data in the presentation space; and a calculation unit 425 configured to calculate, based on the local model and the difference information, an element contribution degree ci,f representing a magnitude of an effect that each element of the features of the neighbor data exerts on the latent distance.

Privacy-preserving security screening using facial images is disclosed. The screening may comprise encoding a facial image into a latent vector using a first pre-trained machine learning model, shuffling the latent vector using an invertible transformation in a latent space to generate a shuffled latent vector and converting the shuffled latent vector into a landscape image using a second pre-trained machine learning model. The landscape image may be used to security screening and compared with reference landscape images.

A method for training a model for generating a representation of long-term motion from a text input comprises: training a motion encoder of an autoencoder to compress and map an input motion into a latent representation comprising a sequence of latent vectors in a discrete latent space, each latent vector representing a fixed length of motion; training a quantization module to quantize the latent vectors to a sequence of quantized latent vectors in quantized latent space; and training a motion decoder to reconstruct the quantized sequence as a sequence of single-frame pose representations. A text encoder is trained to predict a latent sequence conditioned on a text input and a duration using the mapped latent representation as a target.

Causal discovery is performed using knowledge graph link prediction. Information from a causal network is transformed into a causal knowledge graph according to a mapping, the causal knowledge graph including a plurality of causal links, wherein each causal link includes a cause entity, a causal relation, an effect entity, and a causal weight indicating a relative strength of causal influence of the cause entity on the effect entity. The causal knowledge graph is converted into embeddings, where the embeddings include a latent vector space representation of the causal knowledge graph. The embeddings are trained using a subset of the causal links of the causal knowledge graph. The embeddings are used for causal discovery to predict additional causal links of the causal knowledge graph.

The invention provides a semantic-driven real scene video super-division method and system, and belongs to the technical field of video data. Understanding the obtained low-resolution video, and converting the low-resolution video into a hierarchical semantic representation form; channel self-attention and channel mutual attention are alternately used for the obtained low-resolution video to extract intra-frame features and inter-frame features, fusion is carried out, aligned features are obtained through motion compensation operation, and coarse-grained video restoration is carried out according to the aligned features; adding Gaussian noise to the coarse-grained restored video to obtain an intermediate state of a current diffusion step, converting the intermediate state to a semantic equivalent space, and decoding the potential vector to generate a refined frame; and performing deep convolution fusion on the coarse-grained restored video and the refined frame to generate a final super-resolution reconstruction frame. According to the method, the reconstruction fidelity and reality sense of the video in the complex degraded scene are improved based on hierarchical semantic representation, and the definition and reality sense of texture details in the reconstructed video are enhanced.

The invention relates to the technical field of civil engineering structure calculation and artificial intelligence crossing, in particular to a load displacement curve prediction method for a CFRP reinforced concrete filled steel tubular column, which comprises the following steps: acquiring load displacement curves and corresponding input parameters of the CFRP reinforced concrete filled steel tubular column under different parameter combinations, and constructing a training database; constructing a load displacement curve prediction model by taking a conditional variation auto-encoder as a framework; for the real-time state parameters of the target CFRP reinforced concrete filled steel tubular column, N times of sampling is carried out from prior distribution of a potential space to obtain N potential vectors, and the N potential vectors and the corresponding real-time state parameters are input into a decoder for prediction to obtain a curve cluster; and averaging the curve clusters to obtain a final prediction curve. According to the method, the variational auto-encoder is combined with the deep neural operator, so that a continuous and smooth whole-process load displacement curve can be directly predicted and generated, and the prediction precision and integrity are greatly improved.

One or more aspects of a method, apparatus, and non-transitory computer readable medium include obtaining an input latent vector for an image generation network and a target lighting representation. A modified latent vector is generated based on the input latent vector and the target lighting representation, and an image generation network generates an image based on the modified latent vector using.