Abstract: A method including determining neighboring items from among sample items based on a respective embedding distance between a query image of a query item and a respective image of each of the neighboring items. The each of the sample items can comprise the respective image and at least one respective item label. The method further can include determining a respective normalized weight for each of the neighboring items based on the respective embedding distance between the query image and the respective image of the each of the neighboring items. The method also can include determining a query item label of the query item based on a weighted majority vote by the neighboring items via the respective normalized weight for the each of the neighboring items. Other embodiments are disclosed.

Abstract: Provided are a method and apparatus for translating words in a picture, an electronic device, and a storage medium. The method includes: recognizing words embedded in a target picture to obtain at least one text line, each of which corresponds to one line of words; perform paragraph combination on the at least one text line to obtain at least one text paragraph; translating the at least one text paragraph into at least one target text paragraph in a specified language; and replacing the words in the target picture with the at least one target text paragraph.

Abstract: An example system includes a processor to receive, a randomly generated alpha-map, a pair of training images, and a pair of training texts associated with the pair of training images. The processor is to generate a blended image based on the randomly generated alpha-map and the pair of training images. The processor is to train a visual language grounding model to separate the blended image into a pair of heatmaps identifying portions of the blended image corresponding to each of the training images using a separation loss.

Abstract: The present application relates to an image processing method and system. The method may include: acquiring a sequence of input images containing a target object; and performing multi-resolution fusion on the sequence of input images to generate a single fused image, where pixels of the fused image may include a pixel at a corresponding position of an input image in the sequence of input images, and each pixel of the fused image containing the target object may include a pixel at a corresponding position of an input image in the sequence of input images in which part of the target object is focused.

Abstract: The present disclosure provides methods, apparatuses, devices and storage medium for predicting correlation between objects. The method can include: detecting a first object, a second object, and a third object involved in a target image, wherein the first object and the second object represent different body parts, and the third object indicates a body object; determining a joint bounding box surrounding the first object, the second object, and the third object; and predicting correlation between the first object and the second object based on a region corresponding to the joint bounding box in the target image.

Abstract: Embodiments are disclosed for a method for machine-learning model accuracy. The method includes generating prediction training data based on training predictions and corresponding probabilities of the training predictions. A classifier of a machine-learning model generates the training predictions. The method also includes training a prediction accuracy model to determine whether the training predictions generated by the machine-learning model are correct. Additionally, the method includes generating predictions in response to corresponding client transactions for the machine-learning model. Further, the method includes determining whether the predictions are accurate using the prediction accuracy model. Also, the method includes providing client predictions corresponding to the client transactions based on the determination.

Abstract: Embodiments described herein embodiments described herein provide Contrastive Attention-Supervised Tuning (CAST), a training method to fix the visual grounding ability of contrastive SSL methods based on a data augmentation strategy using unsupervised saliency maps. In addition to the contrastive loss that encourages the model to pick the crop that comes from the corresponding image, CAST provides an explicit grounding supervision through a Grad-CAM based attention loss that enforces models to look at the specified object of interest that is common across different crops when making this decision. A new geometric transform is introduced for randomly cropping different views from an input image based on certain constraints derived from a saliency map.

Abstract: A laparoscopic image smoke removal method based on a generative adversarial network, and belongs to the technical field of computer vision. The method includes: processing a laparoscopic image sample to be processed using a smoke mask segmentation network to acquire a smoke mask image; inputting the laparoscopic image sample to be processed and the smoke mask image into a smoke removal network, and extracting features of the laparoscopic image sample to be processed using a multi-level smoke feature extractor to acquire a light smoke feature vector and a heavy smoke feature vector; and acquiring, according to the light smoke feature vector, the heavy smoke feature vector and the smoke mask image, a smoke-free laparoscopic image by filtering out smoke information and maintaining a laparoscopic image by using a mask shielding effect. The method has the technical effects of robustness and ability of being embedded into a laparoscopic device for use.

Abstract: An anomaly detection apparatus according to an embodiment of the present disclosure includes: a global tree structure creation unit configured to create a global tree structure for dividing a plurality of data pieces into a plurality of groups, a local tree structure creation unit configured to create a local tree structure for further dividing the data pieces divided into the plurality of groups for each of the plurality of groups, and a score calculation unit configured to calculate a score indicating an anomaly level of the plurality of data pieces using a depth from a root node to a leaf node of the local tree structure.

Abstract: Methods for creating a knowledge graph for a video are disclosed. Aspects include obtaining the video, processing the video to extract audio information and video information, and storing the extracted audio information and video information with a timestamp corresponding its occurrence in the video. Aspects also include creating a plurality of groups of the extracted audio information and video information based at least in part on the timestamps and extracting two or more keywords from each of the plurality of groups. Aspects further include identifying a relationship between the two or more keywords based on the extracted audio information and video information and creating a graph having a plurality of nodes and a plurality of links that connect a pair of nodes of the plurality of nodes. Each node corresponds to an extracted keyword and each link corresponds to the identified relationship between the pair of nodes.

Abstract: Systems and techniques for automatic digital parameter adjustment are described that leverage insights learned from an image set to automatically predict parameter values for an input item of digital visual content. To do so, the automatic digital parameter adjustment techniques described herein captures visual and contextual features of digital visual content to determine balanced visual output in a range of visual scenes and settings. The visual and contextual features of digital visual content are used to train a parameter adjustment model through machine learning techniques that captures feature patterns and interactions. The parameter adjustment model exploits these feature interactions to determine visually pleasing parameter values for an input item of digital visual content. The predicted parameter values are output, allowing further adjustment to the parameter values.

Abstract: Provided is an image determination device. The image determination device is provided with: feature extractors which output, on the basis of an image to be examined, each piece of feature data indicating a specific feature of the image; a determiner which outputs, on the basis of the feature data output from the extractors, output data indicating the determination result pertaining to the image; and a training part which trains the determiner so as to output, output data indicating the label data associated with the training image on the basis of the feature data output when the training image is input to the extractors, wherein the training part further trains, by using new training data, the determiner so that the output data indicating the label data associated with the image is output by the determiner.

Abstract: A method includes accessing a web-based property over a network; storing a plurality of images or videos from the web-based property and associations between the plurality of images or videos and a target audience identifier responsive to the web-based property having a stored association with the target audience identifier; retrieving the plurality of images or videos from the database responsive to each of the plurality of images or videos having stored associations with the target audience identifier; executing a neural network to generate a performance score for each of the plurality of images or videos; calculating a target audience benchmark; executing the neural network to generate a first performance score for a first image or video and a second performance score for a second image or video; comparing the first performance score and the second performance score to the benchmark; and generating a record identifying the first image or video.

Abstract: Disclosed herein are methods, systems, and computer-readable media for generating an image corresponding to a text input. In an embodiment, operations may include accessing a text description and inputting the text description into a text encoder. The operations may include receiving, from the text encoder, a text embedding, and inputting at least one of the text description or the text embedding into a first sub-model configured to generate, based on at least one of the text description or the text embedding, a corresponding image embedding. The operations may include inputting at least one of the text description or the corresponding image embedding, generated by the first sub-model, into a second sub-model configured to generate, based on at least one of the text description or the corresponding image embedding, an output image. The operations may include making the output image, generated by the first second sub-model, accessible to a device.

Abstract: A learning data generation apparatus includes an object extraction unit configured to extract an object image from an image; a classification evaluation unit configured to evaluate the object possireimage based on a learned model, and to calculate reliability indicating a degree of posibility that the object image is classified as a candidate label; a classification determination unit configured to, if the reliability is smaller than a first threshold and equal to or larger than a second threshold which is smaller than the first threshold, associate a temporary label different from the candidate label with the object image; and a learning data generation unit configured to generate learning data based on the object image that is associated with the temporary label.

Abstract: In various embodiments, a computer-implemented method of training a neural network for relighting an image is described. A first training set that includes source images and a target illumination embedding is generated, the source images having respective illuminated subjects. A second training set that includes augmented images and the target illumination embedding is generated, where the augmented images corresponding to the source images. A first autoencoder is trained using the first training set to generate a first output set that includes estimated source illumination embeddings and first reconstructed images that correspond to the source images, the reconstructed images having respective subjects that are i) from the corresponding source image, and ii) illuminated based on the target illumination embedding.

Abstract: A system for generating and interacting with a virtual model of a physical entity is disclosed. The system includes a processing subsystem which includes an input module which receives preference(s) and data corresponding to parameter(s). The processing subsystem also includes a model generation module which performs a fusion operation on the data, thereby generating point cloud data and generates a crude virtual model corresponding to the physical entity. The processing subsystem also includes a model improvement module which performs cleaning and optimization of the crude virtual model, generates the virtual model, identifies multiple objects associated with the physical entity, and generates sub-virtual model(s) for the multiple objects.

Abstract: The present invention makes an efficient diagnosis of an object. A general server 3 of a diagnostic assistance system 1 provides data (a large volume of photographic images) concerning an object managed by each client, to said client and a plurality of analysts and experts, and allows such diagnosers to diagnose the object and enables sharing of diagnosis results among the diagnosers. The plurality of analysts each partially contribute in viewing the large volume of photographic images so as to find an abnormal site in the object. The client and the experts then conduct a more detailed diagnosis on the abnormal site discovered by the analysts. An AI server 5 of the diagnostic assistance system 1 creates training data from the diagnosis results provided by the plurality of diagnosers, performs machine learning on the diagnosis results, and carries out automated diagnosis using the learned method.

Abstract: Described is a system for improving generalization of an agent, such as an autonomous vehicle, to unanticipated environmental changes. A set of concepts from the agent's experiences of an environment are extracted and consolidated into an episodic world model. Using the episodic world model, a dream sequence of prospective simulations, based on a selected set of concepts and constrained by the environment's semantics and dynamics, is generated. The dream sequence is converted into a sensor data format, which is used for augmented training of the agent to operate in the environment with improved generalization to unanticipated changes in the environment.

Abstract: Improved multi-modal machine learning networks integrate computer vision systems with language models. In certain embodiments, a computer vision system analyzes at least one image to generate a computer vision output. The language model generates an output based, at least in part, on a consideration of the computer vision output. The outputs of the language model can be generated by jointly considering textual information learned by the language model and visual content extracted by the computer vision system, thereby significantly improving the accuracy, breadth, and comprehensiveness of the outputs.

Abstract: An image generation system is provided to: receive a 3D CAD (computer aided design) model comprising 3D model images of a target object; based on the 3D CAD model of the target object, generate augmented CAD models of the target object comprising data sets, each data set respectively corresponding to an associated one of a plurality of attribute classes, each data set comprising a plurality of 2D model images; input the data sets into a generative adversarial network comprising generators respectively corresponding to the plurality of attribute classes and discriminators respectively corresponding to the generators; generate synthetic photorealistic images of the target object using the generators, the synthetic photorealistic images including attributes in accordance with the data sets corresponding to the plurality of attribute classes; and output the synthetic photorealistic images of the target object.

Abstract: An information processing method includes: obtaining sensing data; determining a synthesis region in the sensing data in which recognition target data is to be synthesized with the sensing data; generating composite data by synthesizing the recognition target data into the synthesis region, the recognition target data having same or similar characteristics perceived by a human sensory system as the sensing data; obtaining recognition result data by providing the composite data to a model which has been trained using machine learning to recognize a recognition target; making a second determination based on the recognition result data and reference data including at least the synthesis region, the second determination being to determine whether to make a first determination, the first determination being to determine training data for the model based on the composite data; and making the first determination when it is determined in the second determination to make the first determination.

Abstract: A customizable recognition system with at least one processor to process the audio/video input to determine a control command for accessibility functions of a computer or gaming application. The customized recognition engine has a classifier for each different input type for the different types of speech or gestures. The classifier stored with a link or indication of a user identifier. The interface is configured to provide the control commands to a computer application, gaming application, or a laptop, or an access technology device.

Abstract: An online concierge system uses a machine learning click through rate model to select promoted items based on user embeddings, item embeddings, and search query embeddings. Embeddings obtained by an embedding model may be used as inputs to the click through rate model. The embedding model may be trained using different actions to score the strength of a customer interaction with an item. For example, a customer purchasing an item may be a stronger signal than a customer placing an item in a shopping cart, which in turn may be a stronger signal than a customer clicking on an item. The online concierge system generates a ranking of candidate promoted items based on the search query and using the click through rate model. Based on the ranking, the online concierge system displays promoted items along with the organic search results to the customer.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing images or features of images using an image classification system that includes a batch normalization layer. One of the systems includes a convolutional neural network configured to receive an input comprising an image or image features of the image and to generate a network output that includes respective scores for each object category in a set of object categories, the score for each object category representing a likelihood that that the image contains an image of an object belonging to the category, and the convolutional neural network comprising: a plurality of neural network layers, the plurality of neural network layers comprising a first convolutional neural network layer and a second neural network layer; and a batch normalization layer between the first convolutional neural network layer and the second neural network layer.

Abstract: Aspects relate to systems and methods for parsing and correlating solicitation video content. An exemplary system includes a computing device configured to receive a solicitation video related to a subject, where the solicitation video includes at least an image component and at least an audio component, where the audio component includes audible verbal content related to at least an attribute of the subject, transcribe at least a keyword as a function of the audio component, and associate the subject with at least a job description as a function of the at least a keyword.

Abstract: According to examples, a method comprises obtaining a digital representation of a three-dimensional object, mapping the digital representation on to a sphere of a pre-determined radius and generating a descriptor of the digital representation based on a spherical harmonic decomposition of an output of the mapping.

Abstract: At least one storage stores a dictionary having information corresponding to crowd state images. At least one processor extracts rectangular regions, a size of the rectangular regions is predetermined, from the given image, and recognizes crowd states in the extracted rectangular regions based on the dictionary. The dictionary is acquired by machine learning by use of a plurality of pairs of crowd state images and training label for the crowd state image.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using comfort scales to assess the performance of autonomous vehicles. One of the methods includes receiving data representing a traffic encounter between a vehicle and a pedestrian. A plurality of comfort scale ratings of the encounter assigned by a rater belonging to a first rater pool are received. An input element is generated for a rating transformation model configured to predict how comfort scale ratings assigned by a particular rater pool would have been assigned by a representative rater belonging to a reference rater pool. An inference pass is performed over the rating transformation model using the input element to obtain a plurality of transformed comfort scale ratings for the reference rater pool.

Abstract: A system includes a processor and a hardware accelerator coupled to the processor. The hardware accelerator includes data analysis elements configured to analyze a data stream based on configuration data and to output a result, and an integrated circuit device that includes a DMA engine that writes input data to and read output data from the data analysis elements, one or more preprocessing cores that receive the input data from the DMA engine prior to the DMA engine writing the input data to the one or more data analysis elements and perform custom preprocessing functions on the input data, and one or more post-processing cores that receive the output data from the DMA engine after the output data is read from the data analysis elements but prior to the output data being output to the processor and perform custom post-processing functions on the output data.

Abstract: The present disclosure is directed to extracting text from form-like documents. In particular, a computing system can obtain an image of a document that contains a plurality of portions of text. The computing system can extract one or more candidate text portions for each field type included in a target schema. The computing system can generate a respective input feature vector for each candidate for the field type. The computing system can generate a respective candidate embedding for the candidate text portion. The computing system can determine a respective score for each candidate text portion for the field type based at least in part on the respective candidate embedding for the candidate text portion. The computing system can assign one or more of the candidate text portions to the field type based on the respective scores.

Abstract: A method and system for image-based region detection. Transformation matrices are computed by performing image registration between a target image and each of one or more reference images. Each transformation matrix is for transforming each of the reference images into a coordinate system of the target image. An optimal reference image is selected from among the reference images based on similarity measures between the target image and each reference image. The transformation matrix of the selected reference image is applied to a reference map associated with the reference image in order to generate a target map for the target image. The target map includes region labels indicating regions shown in the target image.

Abstract: Systems and methods for processing image data representing a document to remove deformations contained in the document are disclosed. A system may include one or more memory devices storing instructions and one or more processors configured to execute the instructions. The instructions may instruct the system to provide, to a machine learning system, a training dataset representing a plurality of documents containing a plurality of training deformations. The instructions may also instruct the system to use the machine learning system to process image data representing a target document containing a target document deformation. The machine learning system may generate restored image data representing the target document with the target document deformation removed. The instructions may further instruct the system to provide the restored image data to at least one of a graphical user interface, an image storage device, or a computer vision system.

Abstract: A microscopy system comprises a microscope which can record at least one microscope image and a computing device which comprises a trained image processing model set up to calculate an image processing result based on the at least one microscope image. A method for verifying a trained image processing model, which can be performed by the computing device, can include receiving a validation image and an associated target image; entering the validation image into the trained image processing model, which calculates an output image therefrom; entering image data based on at least the output image and the associated target image into a trained verification model which is trained to calculate an evaluation that indicates a quality that depends on the image data for entered image data; and calculating an evaluation by the trained verification model based on the entered image data.

Abstract: Described herein are systems and methods that search videos and other media content to identify items, objects, faces, or other entities within the media content. Detectors identify objects within media content by, for instance, detecting a predetermined set of visual features corresponding to the objects. Detectors configured to identify an object can be trained using a machine learned model (e.g., a convolutional neural network) as applied to a set of example media content items that include the object. The systems and methods describe techniques that search videos and media content to determine the presence of unknown objects, generate novel detectors trained to identify the unknown objects, and apply the novel detectors to historical media content to identify previous appearances of the unknown objects.

Abstract: An embodiment of the present disclosure provides an artificial intelligence apparatus for generating training data including a memory configured to store an artificial intelligence model, an input interface including a microphone or a camera, and a processor configured to receive, via the input interface, input data, generate an inference result corresponding to the input data by using the artificial intelligence model, receive feedback corresponding to the inference result, determine suitability of the input data and the feedback for updating the artificial intelligence model, and generate training data based on the input data and the feedback if the input data and the feedback are determined as data suitable for updating of the artificial intelligence model.

Abstract: An information management apparatus comprises a communication unit configured to communicate with a plurality of external apparatuses having learning functions, and a control unit configured to control the communication with the plurality of external apparatuses performed by the communication unit. The control unit, if supervisory data generated when a predetermined external apparatus executes a learning function is received from the predetermined external apparatus via the communication unit, selects, from among the plurality of external apparatuses, an external apparatus, other than the predetermined external apparatus, with which the supervisory data is to be shared, and performs control so that the supervisory data is transmitted to the selected external apparatus.

Abstract: A multiple instance learning device for analyzing 3D images, comprises a memory in which a multiple instance learning model is stored and at least one processor electrically connected to the memory, wherein the multiple instance learning model comprises a convolution block configured to derive a feature map for each of 2D instances of a 3D image inputted to the multiple instance learning model, a spatial attention block configured to derive spatial attention maps of the instances from the feature maps derived from the convolution block, an instance attention block configured to receive a result of combining the feature maps and the spatial attention maps and derive an attention score for each instance, and derive an aggregated embedding for the 3D image by aggregating embeddings of the instances according to the attention scores.

Abstract: Improved multi-modal machine learning networks integrate computer vision systems with language models. In certain embodiments, a computer vision system analyzes at least one image to generate a computer vision output. The language model generates an output based, at least in part, on a consideration of the computer vision output. The outputs of the language model can be generated by jointly considering textual information learned by the language model and visual content extracted by the computer vision system, thereby significantly improving the accuracy, breadth, and comprehensiveness of the outputs.

Abstract: The invention relates to a control device (1) for a vehicle for determining the perceptual load of a visual and dynamic driving scene. The control device is configured to: ?receive a sensor output (101) of a sensor (3), the sensor (3) sensing the visual driving scene, ?extract a set of scene features (102) from the sensor output (101), the set of scene features (102) representing static and/or dynamic information of the visual driving scene, ?determine the perceptual load (104) of the set of extracted scene features (102) based on a predetermined load model (103), the load model (103) being predetermined based on reference video scenes each being labelled with a load value, ?map the perceptual load to the sensed driving and ?determine a spatial and temporal intensity distribution of the perceptual load across the sensed driving scene. The invention further relates to a vehicle, a system and a method.

Abstract: An example apparatus for mining multi-scale hard examples includes a convolutional neural network to receive a mini-batch of sample candidates and generate basic feature maps. The apparatus also includes a feature extractor and combiner to generate concatenated feature maps based on the basic feature maps and extract the concatenated feature maps for each of a plurality of received candidate boxes. The apparatus further includes a sample scorer and miner to score the candidate samples with multi-task loss scores and select candidate samples with multi-task loss scores exceeding a threshold score.

Abstract: A process and a system for creating a visual guide for developing training data for a classification of image, where the training data includes images tagged with labels for the classification of the images. A processor may prompt a user to define a framework for the classification. For an initial set of images within the training data, qualified human classifiers are prompted to locate the images within the framework and to tag the images with labels. The processor determines whether the tagged images have consistent labels, and, if so, the processor adds images to the training data. The processor may add the images by providing a visual guide, the visual guide including tagged images arranged according to their locations within the framework their labels, and prompting human classifiers to tag the additional images with labels for the classification, according to the visual guide.

Abstract: In various examples, object detections of a machine learning model are leveraged to automatically generate new ground truth data for images captured at different perspectives. The machine learning model may generate a prediction of a detected object at the different perspective, and an object tracking algorithm may be used to track the object through other images in a sequence of images where the machine learning model may not have detected the object. New ground truth data may be generated as a result of the object tracking algorithms outputs, and the new ground truth data may be used to retrain or update the machine learning model, train a different machine learning model, or increase the robustness of a ground truth data set that may be used for training machine learning models from various perspectives.

Abstract: A method for performing a crowdsourcing task is provided. The method executed on a computing device comprises generating a crowdsourcing task to take at least one image of an original face photo; sending the generated crowdsourcing task to a crowdsourcing server to publish it thereon, the crowdsourcing server communicating with at least one user device registered in the crowdsourcing server; receiving at least one confirmation from the crowdsourcing server, each confirmation corresponding to the crowdsourcing task accepted by a particular user with a corresponding one of the registered user devices; providing an original face photo for each user corresponding to one of the received confirmations; sending, through the crowdsourcing server, the original face photo to a particular user device corresponding to the user; and receiving at least one image from the user device, the received images being taken for the original face photo in accordance with the accepted crowdsourcing task.

Abstract: An image description information generation method includes obtaining a to-be-processed target image, and inputting the target image into a target-image description information generation network. The target-image description information generation network is a generation network that is obtained by performing adversarial training using a plurality of sample images and that is configured to generate image description information, the adversarial training is training an initialized image description information generation network and an initialized discriminative network alternately, and the discriminative network is configured to discriminate an output result of the image description information generation network. The method also includes according to the output result of the target-image description information generation network, generating target-image description information used for describing the target image.

Abstract: In some implementations, a device may obtain a set of input images, of an object of interest, that includes images that have a plain background. The device may obtain a set of background images that includes images associated with the object of interest. The device may generate, for an input image, a first modified image of the input image that removes a plain background of the input image. The device may generate, for the input image, a second modified image of the input image that is based on the first modified image and a background image. The device may generate, for the input image, a training image that includes an indication of a location of the object of interest depicted in the training image. The device may provide the training image to a training set, that includes a set of training images, for a computer vision model.

Abstract: Various embodiments are described that relate to classification of an unknown object. A time series signal associated with an unknown object can be obtained from a sensor. The time series signal can be subjected to a transform set, such as a Fourier transform and a discrete cosine transform, to produce a transform outcome. Based, at least in part, on the transform outcome, the unknown object can be classified.

Abstract: Methods and systems for performing concept-based adversarial generation with steerable and diverse semantics. One system includes an electronic processor configured to access an input image. The electronic processor is also configured to perform concept-based semantic image generation based on the input image. The electronic processor is also configured to perform concept-based semantic adversarial learning using a set of semantic latent spaces generated as part of performing the concept-based semantic image generation. The electronic processor is also configured to generate an adversarial image based on the concept-based semantic adversarial learning. The electronic processor is also configured to test a target model using the adversarial image.

Abstract: Embodiments are disclosed for generative image congealing which provides an unsupervised learning technique that learns transformations of real data to improve the image quality of GANs trained using that image data. In particular, in one or more embodiments, the disclosed systems and methods comprise generating, by a spatial transformer network, an aligned real image for a real image from an unaligned real dataset, providing, by the spatial transformer network, the aligned real image to an adversarial discrimination network to determine if the aligned real image resembles aligned synthetic images generated by a generator network, and training, by a training manager, the spatial transformer network to learn updated transformations based on the determination of the adversarial discrimination network.

Abstract: Methods, systems, and apparatus for detecting correlated objects involved in images are provided. In one aspect, a method includes: detecting a face object, a preset body part object, and a hand object involved in an image, performing a respective correlation prediction on every two of the face object, the preset body part object, and the hand object to obtain first, second, and third correlation prediction results, segmenting the image to determine at least one body object involved in the image to determine a first body object to which the face object belongs and a second body object to which the preset body part object belongs, adjusting the first correlation prediction result based on the first body object and the second body object, and determining correlated objects involved in the image according to the adjusted first correlation prediction result and the second and third correlation prediction results.