416 results about "Image retrieval" patented technology

The invention relates to an enhanced LLM-RAG multi-hop question and answer method based on logic tree reasoning, and belongs to the technical field of new-generation information, and the method comprises the following steps: inputting a multi-hop question and answer question into a computer system; the computer system calls a pre-training large language model LLM, the multi-hop question-answer question is decomposed into a hierarchical logic tree in a recursive mode, and each node of the logic tree comprises a sub-question and a corresponding hypothesis answer; performing image retrieval from a structured knowledge source Wikidata and performing text retrieval from an unstructured knowledge source Wikipedia on the basis of each node sub-question and the hypothesis answer to obtain corresponding evidence; traversing the logic tree, verifying the consistency between the hypothetical answer of each node and the evidence through LLM, if the contradiction exists, reconstructing the corresponding sub-tree, and dynamically correcting the reasoning path; and integrating the verified logic tree node information, and outputting an accurate answer to the multi-hop question and answer question.

The invention provides a cross-modal joint contrast learning method and device and electronic equipment, and the method comprises the steps: constructing a sample data set which covers a plurality of task types, such as a text retrieval image, an image retrieval text, a text retrieval text, an image retrieval image, an image-text joint retrieval image and an image-text joint retrieval text; and generating prompt words for identifying task types for each task sample, splicing the prompt words with sample data to form task input, and determining modal types of a retrieval object and a retrieval target. A retrieval object and a retrieval target are respectively input into encoders of corresponding modes to extract features, the features are mapped to the same semantic space through a unified projection layer to obtain an embedded vector, and the parameters of the encoders and the projection layer are optimized by utilizing a contrast learning loss function based on the similarity of the retrieval object and the retrieval target, so that multi-task unified training is realized. Various cross-modal retrieval tasks can be supported in a unified semantic space at the same time, and the overall retrieval effect is improved on the premise of ensuring multi-task performance balance.

The invention discloses a cross-view image geographic positioning method based on dynamic threshold pseudo tag self-training learning, and the method specifically comprises the following steps: introducing a difficult sample feature mining method, dynamically adjusting the loss weight of a sample according to the change of similarity, and building a dynamic difficult sample triple loss model; the method comprises the following steps: dynamically adjusting a confidence threshold value of a sample by adopting an index moving average weighting method, iteratively training and screening an unlabeled sample, namely a pseudo label, establishing a pseudo label self-training mechanism of a dynamic threshold value, mining and utilizing non-paired data, and solving the problem of high manual labeling cost; a reference image most similar to a query image is found through image retrieval, and the offset of a query position is predicted. Experiments on CVUSA and CVACT data sets show that as the distance threshold increases, the accuracy of the cross-view image geographic positioning method based on dynamic threshold pseudo tag self-training learning presents a stable rising trend, and the cross-view image geographic positioning method based on dynamic threshold pseudo tag self-training learning is superior to other methods under the same threshold condition.

The invention provides an image searching method, which comprises the following steps of: when a new label is added to an image in a first image database, acquiring a first semantic feature of the new label; based on the first semantic feature, performing image retrieval in a first image database to obtain a first retrieval result and a confidence coefficient of the first retrieval result; if the confidence coefficient is smaller than the preset confidence coefficient, performing secondary retrieval in the first retrieval result based on the existing tag in the first retrieval result to obtain a second retrieval result; based on the second retrieval result, adding the newly added tag into an image corresponding to the secondary retrieval result to obtain a second image database; and after an image search instruction of the user is obtained, performing image retrieval in the second image database based on the image search instruction. The problems that in the database maintenance process, a new label needs to be added, the workload of adding the label again to the data is increased while the new label is added, and the database is difficult to maintain in an existing method are solved.

The disclosure relates to systems and methods automatically extracting an image and related image components, computationally determining an understanding of the image, and generating mathematical vector embeddings via sentence encoders based on the computationally determined understanding. The mathematical vector embeddings may be used for semantic image retrieval that enables image searching based on a semantic understanding of input images and / or input text. The mathematical vector embeddings may be used for training and executing generative Artificial Intelligence (AI) models to create new content that includes retrieved images and / or generate new images.

The invention relates to the technical field of power grid image retrieval, and discloses a power distribution network inspection image rapid retrieval matching method and system, and the method comprises the steps: obtaining a to-be-retrieved inspection image of power distribution network equipment, and extracting the equipment structure features of the inspection image through a hierarchical convolutional network; quantifying a surface texture attenuation index of the power distribution network equipment through fractal dimension based on the equipment structure characteristics; performing mapping relation coupling on the surface texture attenuation index and a space coordinate of an equipment connecting piece to generate a dynamic feature coding sequence containing an equipment structure topological relation; performing time sequence consistency matching on the dynamic feature coding sequence and a pre-constructed reference image library, and aligning an equipment aging track through a dynamic time warping algorithm to generate a similarity sorting result; according to the method, the problems that effective features cannot be extracted during retrieval matching and the retrieval precision is low are solved.

A method, apparatus, non-transitory computer readable medium, and system for media processing include obtaining a text prompt describing content, generating, using a multi-modal encoder, a text embedding based on the text prompt, and obtaining an image depicting the content based on the text embedding. The multi-modal encoder is trained to encode image descriptions based on a similarity between a caption of a training image and a paraphrase of the caption.

The invention relates to the technical field of computer vision processing, in particular to an automatic defect positioning method based on a BIM virtual image. The method comprises the following steps: establishing a high-rise building model, and making a data set integrating an illumination condition and a full view angle; calculating camera parameters; performing cross-modal image retrieval: performing targeted optimization on the basis of a classical ResNet architecture, and constructing a backbone network structure suitable for a building image retrieval task; initializing position attitude estimation based on matching; and correcting the camera position posture. According to the method, a defect fixed frame based on a building BIM virtual image is provided, a building image data set BIM-Vision based on Revit is constructed, rich visual angles and illumination condition setting are achieved, accurate camera position postures and 3D labels between beam columns are provided, high-quality basic data support is provided for building visual research, and the method has the advantages of being high in practicability and high in practicability. And during inspection, accurate positioning of defect positions and component association can be completed only by shooting a field image, so that the field operation process is greatly simplified.

The invention relates to the technical field of visual position recognition, and particularly discloses a multi-modal visual position recognition reordering method and system based on guidance, and the method comprises the steps: obtaining a query image, and retrieving a plurality of candidate images based on a pre-trained visual basic model and the query image; constructing a composite multi-modal prompt object, wherein the composite multi-modal prompt object comprises an image pair formed by the query image and the current candidate image, and an instruction text used for guiding a multi-modal large language model to perform visual comparison; outputting a structured similarity judgment result, wherein the result comprises a quantitative similarity score; and sorting based on the similarity scores corresponding to all the candidate images, and determining the candidate image with the highest score as an optimal matching result. Through combination of guiding type prompt engineering and structured output, an intermediate text generation link is avoided fundamentally, and the calculation efficiency is improved while the fidelity of all original visual information is reserved.

The invention discloses a distributed medical image retrieval and hierarchical storage management system and method. The system comprises an interface processing module, a distributed image storage cluster, a routing forwarding module, a metadata management module and a storage management module. The interface processing module receives the image retrieval request, generates a routing identifier for positioning the storage position of a target medical image file and sends the routing identifier to the routing forwarding module; and the routing forwarding module queries a mapping relationship between the medical image file and the storage node and / or the storage hierarchy based on the routing identifier, determines a first storage node and forwards the request, so that the first storage node returns a target medical image file. And the storage management module obtains the access statistical information, determines a target storage level and / or a target storage node, migrates the medical image file, and triggers updating of the mapping relation after migration is completed. Therefore, cross-node accurate positioning and unified retrieval are realized, positioning consistency is kept after file migration, and hierarchical storage management based on access conditions is supported.

The invention discloses an image retrieval method and device, and relates to the technical field of knowledge distillation, and the method comprises the steps: taking a query text input by a user as a retrieval, and screening and outputting a target image consistent with text content description from a preset image library through text semantic comprehension and an image feature matching algorithm, the technical problems that in the related technology, it is difficult to deeply mine core knowledge such as fine-grained probability distribution and coarse-grained structural features of an attention mechanism in the image retrieval process, a distillation process cannot cover full-dimensional knowledge of an attention level, and the integrity of knowledge migration is insufficient are solved. The technical effects that through deep correlation matching of the text and the image, the user can be helped to rapidly position the target image through simple text description, the image search operation process is effectively simplified, and efficient and convenient image retrieval and browsing experience are provided for the user are achieved.

An image retrieval method includes acquiring an image retrieval condition, the image retrieval condition comprising a reference image and modification text, the modification text being configured to indicate a modification expectation for the reference image; composing the reference image and the modification text, to obtain an image-text composition; acquiring a plurality of candidate images, and determining, for a candidate image, a first similarity between the candidate image and the image-text composition, and a second similarity between the candidate image and the modification text; and determining at least one target image satisfying the image retrieval condition from the plurality of candidate images with reference to the first similarity and the second similarity.

The invention relates to the field of power grid defect retrieval, in particular to a diffusion model-based deep hash image retrieval method for power grid defect maintenance, which comprises the following steps of: 1, performing fusion coding by inputting text data and image data, and constructing an initial hash code generation model; 2, using a Pair-wise loss function to optimize the distribution of sample pairs in a hash space, introducing a quantization loss function, generating an efficient binary hash code, and generating a high-quality binary hash code; 3, constructing a Hash code-image latent diffusion model, performing diffusion generation by encoding and decoding the Hash code / image to a continuous latent space, enabling the Hash code to correspond to the image in a generative manner, and directly fitting spatial distribution; and 4, defining a loss function of the Hash code-image diffusion model, generating a high-quality Hash code and an image, and obtaining a power grid defect type in a mode of searching images by images. Auxiliary training is carried out through fusion of text features and image features, so that the semantic features understand the images more deeply.

The invention relates to a cyanobacterial bloom remote sensing recognition system based on a multi-source satellite and intelligent threshold correction, and the system comprises an intelligent image retrieval and screening module which is used for calling a multi-source image according to a daily frequency / weekly frequency period and generating a candidate set; the water body mask and preprocessing module is used for cloud removal, cutting, water body extraction and non-target plaque removal; the water bloom index and threshold value module is used for automatically calculating an ABDI or FAI index according to an image source and carrying out peak height proportion threshold value segmentation; the pattern spot extraction and clustering module is used for extracting algal bloom patches, merging pattern spots and removing small patches; and the standardized product generating and pushing module is used for generating and pushing grids, vectors, daily frequency brief reports and weekly frequency thematic reports. According to the invention, daily frequency / weekly frequency automatic and intelligent identification and achievement business output of cyanobacterial blooms can be realized, and the precision and continuity of remote sensing monitoring are significantly improved.

The invention discloses an efficient image retrieval method based on lightweight multi-scale visual Transform and pseudo-space hash. The method comprises the following steps: acquiring an image to be recognized; inputting the preprocessed to-be-recognized image into a pre-trained image retrieval model for processing to generate an image retrieval result, wherein the image retrieval model comprises a Transform backbone network, a multi-scale feature pyramid module and a pseudo-space hash code reconstructor; the method for processing the to-be-recognized image by the image retrieval model comprises the following steps: performing feature extraction on the to-be-recognized image by utilizing a Transform backbone network to obtain multiple layers of basic features; fusing the multiple layers of basic features by using a multi-scale feature pyramid module to obtain high-resolution fusion features; processing the high-resolution fusion feature by using a pseudo-space hash code reconstructor to obtain a retrieval hash code; and performing retrieval based on the retrieval hash code to generate an image retrieval result. According to the method, the complexity is lower, and the compact hash code with discrimination and structure regularity can be generated, so that the accuracy of a retrieval result is improved.

The invention relates to a medical image retrieval method and system based on a DICOM protocol and a storage medium, by setting a unified proxy forwarding service and combining with an image prefetching and caching mechanism linked with a clinical business event, the configuration difference of a back-end heterogeneous PACS system is shielded, single access point connection of a client is realized, and the user experience is improved. The complexity of system deployment and maintenance is reduced; meanwhile, the image data are pre-stored in the cache, so that the retrieval waiting time of the remote image is shortened, and the data access performance and the user experience are improved; and centralized authority authentication and log auditing are realized through the agent layer, so that the controllability and the security of a cross-mechanism data sharing process are guaranteed under the condition of avoiding high cost of centralized storage of full data.

The invention relates to a multi-modal combined image retrieval method fusing fine-grained semantic positioning and optimization generation features, which comprises the following steps: acquiring a reference image, predicting a bounding box of a target object based on the reference image and a positioning keyword, and dividing a background retaining region and a foreground editing region; semantic divergence prompt words of description information are obtained, text features in the semantic divergence prompt words are extracted to serve as guide targets to be used for establishing a composite optimization target function, a gradient descent algorithm is adopted to conduct iterative updating on learnable random noise vectors, and potential visual feature vectors are generated; the foreground editing area is filled with the image data, a background-foreground mixed feature map is generated, the fine-grained visual similarity between the background-foreground mixed feature map and the dense visual feature map is calculated, and the global text feature similarity between global text feature vectors in description information and global visual feature vectors is calculated; and carrying out weighted fusion on the fine-grained visual similarity and the global text feature similarity, and outputting a retrieval result.

The invention discloses an agricultural disease hash retrieval method based on DV stabilization and adaptive feature enhancement, and belongs to the technical field of agricultural disease image retrieval and deep hash, and the method comprises the following steps: 1, obtaining an original disease image as the input of a backbone network; step 2, extracting disease image features by using a backbone network; step 3, based on a self-adaptive feature enhancement module, enhancing the disease image features to obtain self-adaptive image features; step 4, optimizing network parameters through a total loss function containing DV stabilization; step 5, obtaining a binary hash code through the hash layer; and step 6, carrying out Hamming distance sorting on the obtained Hamming codes and the Hamming codes of all the images in the retrieval set obtained by the same method, and returning a plurality of disease images with the Hamming distance smaller than a preset Hamming distance. The agricultural disease image retrieval efficiency and precision are effectively improved.

The invention discloses an unmanned aerial vehicle camera repositioning method and device based on image retrieval and a storage medium. The method comprises the steps of collecting image data and constructing an image database. A global feature extraction module is used for extracting global features to obtain the global features, similarity calculation sorting is carried out on the global features and the global features of the images in the database, and the most similar images are selected to form an image pair. After local feature extraction, relative translation and relative rotation are predicted through a relative pose regression network. And finally, converting the relative pose of the retrieved picture into an absolute pose through a relative pose conversion algorithm. According to the overall method, image feature extraction and pose prediction are realized through a deep learning network, and efficient relocation of the unmanned aerial vehicle camera is realized.

The embodiment of the invention provides a multi-modal image retrieval method and device based on a scene graph, and the method comprises the steps: extracting image features through a built-in visual encoder after receiving a preprocessed image through the powerful understanding capability of a multi-modal large language model, and directly outputting a structured scene graph description through the generation capability of the language model. In this way, in the retrieval process, the model can achieve fine-grained semantic matching through a progressive query retrieval mechanism constructed by query decomposition and subgraphs, and finally high-precision text-image retrieval is achieved.

The invention discloses a privacy protection unsupervised cross-domain image retrieval method and system, and belongs to the technical field of information security. Aiming at the problems of privacy leakage risk and difficult feature alignment in the existing unsupervised cross-domain image retrieval, the invention provides a dual-protection transformation strategy, and realizes local encryption of image features of each data domain by combining orthogonal projection transformation and a differential privacy mechanism. Furthermore, intra-domain discriminant reinforcement learning and differential privacy cross-domain alignment are respectively performed through a probability prototype optimal transmission mechanism, and an intra-domain and cross-domain loss function is constructed based on a contrast learning normal form, so that safe and effective model training is realized. And after training is completed, image similarity calculation is performed by adopting protected feature representation, so that cross-domain image retrieval under privacy protection is realized. According to the method, the accuracy and robustness of cross-domain retrieval are improved, privacy security and model performance are considered, and the method is suitable for an image retrieval task in a multi-party cooperation scene.

The invention provides a visual question and answer method based on a domain self-adaptive retrieval decision, which comprises the following steps of: forming an input triple (x, q, d) comprising an image, a question text and an image description text; feature modal extraction and domain identification; generating an explicit reasoning track and a preliminary answer by using a chain reasoning technology CoT; according to a preset decision rule, judging that the preliminary answer is output as a final answer or enters the next step; based on the input triad (x, q, d) and the reasoning track, image retrieval and text retrieval are executed, and an enhanced knowledge set is generated; using the enhanced knowledge set to generate a final answer through a chain reasoning technology CoT, performing credibility verification, and outputting the final answer or a preset unknown identifier according to a verification result; according to the method, through reasoning-driven adaptive retrieval and multi-modal knowledge reordering, efficient utilization and real-time supplement of external knowledge are realized, and the accuracy and robustness of visual questions and answers are effectively improved.

A method of retrieving visual content includes receiving user input defining an initial search query from a client application. The initial search query and a meta prompt are then delivered to a refined query generating model which is trained to analyze the initial search query to determine user intent and to generate a refined search query based on the initial search query and the meta prompt. The refined search query is delivered to a visual content retrieval model which retrieves aesthetic visual content with reference to a visual content index. Retrieved aesthetic visual content is returned to the client application.

The present disclosure relates to systems, methods, and non-transitory computer readable media that utilize deep learning to identify regions of an image that have been editorially modified. For example, the image comparison system includes a deep image comparator model that compares a pair of images and localizes regions that have been editorially manipulated relative to an original or trusted image. More specifically, the deep image comparator model generates and surfaces visual indications of the location of such editorial changes on the modified image. The deep image comparator model is robust and ignores discrepancies due to benign image transformations that commonly occur during electronic image distribution. The image comparison system optionally includes an image retrieval model utilizes a visual search embedding that is robust to minor manipulations or benign modifications of images. The image retrieval model utilizes a visual search embedding for an image to robustly identify near duplicate images.

The invention discloses a two-dimensional irregular accessory identification method based on depth image retrieval, and belongs to computer vision. The method comprises the following steps: firstly, analyzing the inner and outer contours of an accessory from a CAD drawing, rendering a standard drawing on a canvas with a fixed resolution ratio, and generating a multi-view rendering graph according to a hemisphere discrete pose as simulation data to replace a camera image so as to form a standard-simulation sample pair; then, ResNet18 combined with double attention is adopted to form a twin network, contrastive learning is matched, the same accessory is aggregated in a feature space, different accessories are separated, and vectors are extracted from a standard image in an off-line mode to construct an index; and finally, coding a real image into a query vector in an online stage, performing neighbor retrieval by combining inner product or cosine similarity in a vector library, and returning a Top-K candidate. According to the scheme, both precision and efficiency are considered, and good robustness is achieved for view angle transformation and detail deformation. Under the scene of scarcity of real data or high labeling cost, the training and retrieval performance can be maintained by relying on pre-construction of multi-view simulation data and a vector library.

The invention relates to a combined image retrieval method and system guided by a neighborhood relation based on attributes, and the method comprises the steps: reading training set data in batches, and carrying out the extraction of global features and local features of the training set data; splicing the local features and the global features to form original attribute features of the vision and the text; extracting attribute prototype features; constructing coherent prototype semantics; combining the attribute prototype characteristics of each element in the multi-modal query; for the combined attribute prototype features, modeling is carried out in combination with double relationships, and modeling is carried out on a pairwise relationship and a neighborhood relationship through attribute similarity, so that the measurement learning process is optimized; the COMBINER model generates a corresponding combination feature; similarity scores are solved, and descending order arrangement is carried out; and according to the actual demand, selecting the first K target images as a formal result set so as to complete the combined image retrieval. According to the method, the precision and robustness of combined image retrieval are improved.

The present disclosure relates to an anti-confusion factor based hash image retrieval method, which comprises: obtaining a query image, the query image comprising a reason factor and a confusion factor; inputting the query image into a trained hash network to obtain a hash code expressing the reason factor in the query image; calculating the similarity between the hash code of the query image and hash codes of a plurality of historical images, and taking the historical image corresponding to the hash code with the highest similarity as the image retrieval result. The anti-confusion factor based hash image retrieval method provided by the present disclosure trains a hash network, so that the hash network ignores the confusion factor and focuses on expressing the reason factor in the image in the process of generating a hash code, thereby generating an accurate hash code. Then, by calculating the similarity between the hash codes, an accurate image retrieval result is retrieved.

The application discloses a two-stage COVID-19 antigen detection result discrimination method and device, relates to the field of image processing, and particularly relates to the application of image processing technology in the COVID-19 epidemic prevention and control field, and the method comprises the following steps: step 1, a plurality of quadrilateral regions recording detection results are obtained from a target picture, and a perspective transformation method is used to adjust the visual angle of the quadrilateral regions to obtain a corresponding number of rectangular regions; step 2, a classification network based on deep metric learning is used to identify the rectangular regions to obtain detection results. The application uses a deep metric learning method based on pairs, measures the similarity between sample groups, causes samples of the same type to be close to each other and samples of different types to be separated from each other, and processes fine-grained images at the instance level. The recognition accuracy of tail class samples is improved by adding a memory storage module to optimize the fine-grained image retrieval algorithm.

The invention discloses a rendering graph retrieval analysis method and device, equipment, a medium and a program product. The method comprises the following steps: inputting target rendering graph data into a multi-dimensional analysis processing module for processing to obtain multi-dimensional representation data; the multi-dimensional analysis processing module comprises a feature processing model, a quality evaluation model and a data annotation model, and multi-dimensional representation data comprises normalized image feature vectors, quality scoring information and data annotation information; associating the normalized image feature vector, the quality score information, the data annotation information and the initial rendering image data to obtain feature fusion data; storing the feature fusion data into a vector database, and establishing a target index for the feature fusion data; according to the user query request and the target index, image retrieval operation is executed, and a retrieval analysis result is returned. The rendering graph retrieval efficiency and accuracy can be improved, dependence on manual annotation is remarkably reduced, analysis and evaluation efficiency and objectivity are improved, and the method can be widely applied to the technical field of artificial intelligence.

The invention relates to the technical field of computers, and discloses a document retrieval method, device and equipment, a storage medium and a computer program product.The method comprises the steps that in response to a document retrieval request, the document retrieval request is converted into a problem retrieval vector, and the vector similarity between the problem retrieval vector and a document image retrieval vector is calculated, the document image retrieval vector is obtained by analyzing a document into a document image set and encoding the document image set, and a document retrieval result corresponding to the document retrieval request is generated according to the vector similarity; according to the document retrieval method, the document is analyzed into the document picture set in advance, the document picture set is encoded to obtain the document image retrieval vector, and the document retrieval is performed through the problem retrieval vector and the document image retrieval vector, so that the document retrieval in the form of the document image is realized; therefore, more space structure information, fine-grained information and context information can be reserved, and the document retrieval precision is improved.