Abstract: Embodiments of this disclosure include a method and an apparatus for processing image. The method may include obtaining a to-be-processed image and performing image semantic segmentation on the to-be-processed image to obtain a semantically-segmented image. The semantically-segmented image may include a target region and a non-target region obtained through the semantic segmentation. The method may further include performing pose recognition on the to-be-processed image, to obtain a pose-recognized image recognizing skeletal region. The method may further include fusing the target region and the non-target region of the semantically-segmented image with the skeletal region of the pose-recognized image, to obtain a trimap comprising foreground region, background region, and recognition region. The method may further include generating, according to the to-be-processed image and the trimap, a transparency mask image for separating image from the to-be-processed image.

Abstract: This application relates to a method for constructing a multi-task learning model, an electronic device, and a computer-readable storage medium. The method includes: constructing a search space formed between an input node and a plurality of task nodes by arranging a plurality of subnetwork layers and a plurality of search layers in a staggered manner. A search layer in the plurality of search layers is arranged between two subnetwork layers of the plurality of subnetwork layers. The method includes sampling a path from the input node to each task node of the plurality of task nodes through the search space to obtain a candidate path as a candidate network structure; and training a parameter of the candidate network structure according to sample data, to generate the multi-task learning model for performing a multi-task prediction.

Abstract: A training data acquisition method and device, a server and a storage medium are provided. The training data acquisition method is applied to a classifier and includes the following steps: obtaining an image search target according to an input of a user; providing images to the user according to the image search target, to display the images; and selecting at least one image from the displayed images, and determining a target-classification pair as training data according to the at least one image; where the target-classification pair includes the image search target and an entity-based classification of the at least one image.

Abstract: Embodiments of this disclosure include a method and an apparatus for processing image. The method may include obtaining a to-be-processed image and performing image semantic segmentation on the to-be-processed image to obtain a semantically-segmented image. The semantically-segmented image may include a target region and a non-target region obtained through the semantic segmentation. The method may further include performing pose recognition on the to-be-processed image, to obtain a pose-recognized image recognizing skeletal region. The method may further include fusing the target region and the non-target region of the semantically-segmented image with the skeletal region of the pose-recognized image, to obtain a trimap comprising foreground region, background region, and recognition region. The method may further include generating, according to the to-be-processed image and the trimap, a transparency mask image for separating image from the to-be-processed image.

Abstract: A training data acquisition method and device, a server and a storage medium are provided. The training data acquisition method is applied to a classifier and includes the following steps: obtaining an image search target according to an input of a user; providing images to the user according to the image search target, to display the images; and selecting at least one image from the displayed images, and determining a target-classification pair as training data according to the at least one image; where the target-classification pair includes the image search target and an entity-based classification of the at least one image.

Abstract: A training data acquisition method and device, a server and a storage medium are provided. The training data acquisition method is applied to a classifier and includes the following steps: obtaining an image search target according to an input of a user; providing images to the user according to the image search target, to display the images; and selecting at least one image from the displayed images, and determining a target-classification pair as training data according to the at least one image; where the target-classification pair includes the image search target and an entity-based classification of the at least one image. Thus, more high-quality training data can be obtained, improving the performance of a classifier.

Abstract: Embodiments of the present invention disclose a method and apparatus of generating an image characteristic representation of a query, and an image search method and apparatus. The method of generating an image characteristic representation of a query comprises: acquiring a clicked image set corresponding to a target query based on an image click log of a search user; generating image characteristic vectors corresponding to clicked images in the clicked image set based on image content characteristics of the clicked images; and clustering the clicked images based on the image characteristic vectors, and using a clustering result as an image characteristic representation of the target query.

Abstract: Embodiments of the present invention disclose a method and an apparatus of establishing an image search relevance prediction model, and an image search method and apparatus. The method of establishing an image search relevance prediction model comprises: training a pre-constructed original deep neural network by using a training sample, wherein the training sample comprises: a query and image data, and the original deep neural network comprises: a representation vector generation network and a relevance calculation network; and using the trained original deep neural network as the image search relevance prediction model. The technical solution of the present invention optimizes the existing image search technology, and has stronger capabilities than the prior art as well as various integrations and variations in terms of semantic matching between a query and an image text, semantic matching between a query and image content, click generalization and the like.

Abstract: A training data acquisition method and device, a server and a storage medium are provided. The training data acquisition method is applied to a classifier and includes the following steps: obtaining an image search target according to an input of a user; providing images to the user according to the image search target, to display the images; and selecting at least one image from the displayed images, and determining a target-classification pair as training data according to the at least one image; where the target-classification pair includes the image search target and an entity-based classification of the at least one image. Thus, more high-quality training data can be obtained, improving the performance of a classifier.

Abstract: Embodiments of the present invention disclose a method and an apparatus of establishing an image search relevance prediction model, and an image search method and apparatus. The method of establishing an image search relevance prediction model comprises: training a pre-constructed original deep neural network by using a training sample, wherein the training sample comprises: a query and image data, and the original deep neural network comprises: a representation vector generation network and a relevance calculation network; and using the trained original deep neural network as the image search relevance prediction model. The technical solution of the present invention optimizes the existing image search technology, and has stronger capabilities than the prior art as well as various integrations and variations in terms of semantic matching between a query and an image text, semantic matching between a query and image content, click generalization and the like.

Abstract: Embodiments of the present invention disclose a method and apparatus of generating an image characteristic representation of a query, and an image search method and apparatus. The method of generating an image characteristic representation of a query comprises: acquiring a clicked image set corresponding to a target query based on an image click log of a search user; generating image characteristic vectors corresponding to clicked images in the clicked image set based on image content characteristics of the clicked images; and clustering the clicked images based on the image characteristic vectors, and using a clustering result as an image characteristic representation of the target query. The technical solution of the present invention optimizes the existing image search technology, and can significantly improve the correlation between an image search result and a query entered by a user.

Abstract: Embodiments of the present invention provide a method and apparatus for training an image classifier. The method includes: A. dividing a set of training images for classifier training into a positive-example sample set and at least two negative-example sample sets; B. determining, for each negative-example sample set, a feature set for differentiating the positive-example sample set from the negative-example sample set; and C. performing training using each feature set determined to obtain a classifier. This invention also provides a method and apparatus for image recognition utilizing the image classifier.

Abstract: The embodiments of the present invention provide an image processing method, including: establishing an integral histogram; and calculating a histogram of an arbitrary rectangle area in an image by using the integral histogram. The embodiments of the present invention also provide an image processing system, including an integral histogram establishing unit and a histogram calculating unit. The histogram calculating unit is adapted to calculate a histogram of an arbitrary rectangle area in an image by using an integral histogram established by the integral histogram establishing unit.

Abstract: Embodiments of the present invention provide a method and apparatus for training an image classifier. The method includes: A. dividing a set of training images for classifier training into a positive-example sample set and at least two negative-example sample sets; B. determining, for each negative-example sample set, a feature set for differentiating the positive-example sample set from the negative-example sample set; and C. performing training using each feature set determined to obtain a classifier. This invention also provides a method and apparatus for image recognition utilizing the image classifier.