Abstract: The present disclosure includes a method, apparatus and system for generating a neural network and a non-transitory computer readable storage medium storing instructions. The method comprises: recognizing at least an attribute of an object in a sample image according to a feature extracted from the sample image, using the neural network; determining a loss function value at least according to a margin value determined based on a semantic relationship between attributes, wherein the semantic relationship is obtained from a predefined table at least according to a real attribute and the recognized attribute of the object, wherein the predefined table is composed of the attributes and the semantic relationship between the attributes; updating a parameter in the neural network according to the determined loss function value. When using the neural network generated according the present disclosure, the accuracy of object attribute recognition can be improved.

Abstract: A detection apparatus to extract features from an image; determine the number of candidate regions of the object in the image based on the extracted features, wherein the determined number of the candidate regions is decided by a position and shape of the candidate regions; and to detect the object from the image based on at least the extracted features and the determined number, position and shape of the candidate regions.

Abstract: The present disclosure provides a detection apparatus and method, and image processing apparatus and system. The detection apparatus extracts features from an image, detects objects in the image based on the extracted features; and detects key points of the detected objects based on the extracted features, the detected objects and a pre-obtained key point sets. According to the present disclosure, the whole detection speed can be ensured not to be influenced by the number of objects in the image to be detected while the objects and key points thereof are detected, so as to better meet the requirements of timeliness and practicability of the detection by the actual computer vision task.

Abstract: A recognition apparatus for extracting first features of an object from a region of the object; obtaining second features of the object at least based on confidence information for first attribute of the object; determining sample information from pre-determined sample information based on the obtained second features; and recognizing the first attribute of the object based on the determined sample information.

Abstract: Embodiments of the present disclosure provide a quantization method of feature vector, search method, apparatus and storage medium, wherein the quantization method comprises: setting quantization codewords; selecting, from the set quantization codewords, at least one quantization codeword of which distance from an original feature vector is smaller than first predetermined distance as a quantization codeword subset; determining quantization feature vectors corresponding to the original feature vector by using at least two quantization codewords, wherein at least one quantization codeword from the at least two quantization codewords are in the quantization codeword subset. The search method is a method of performing a search among quantization feature vectors after a quantization using the quantization method.

Abstract: A template determining apparatus including an attribute distribution determination unit configured to determine a distribution of a specific attribute in a plurality of images; and a template determination unit configured to adaptatively determine a template set from the plurality of images according to the determined distribution of the specific attribute of the plurality of images. Where the determined template set will be used for image normalization.

Abstract: An image processing apparatus for extracting features from video frames of a video; and determining, for non-initial video frames, reference information of an object detected in a previous video frame thereof in a corresponding non-initial video frame with respect to object information of the object; and detecting an object from an initial video frame based on the features and detects an object from non-initial video frames based on the features and the determined reference information. The processing time of the object detection processing can be reduced, and the real-time requirements of object detection in the video can be better satisfied.

Abstract: This invention provides a method for extracting a diagnosis object from a medical document comprises: extracting, from an input medical document, body part entities and at least one type of non-body-part entities and the relations between the body part entities and the non-body-part entities; obtaining, for each pair of all possible pairs of the non-body-part entities, a relevance score between two non-body-part entities within one pair, wherein the relevance score is obtained by using the relations between the two non-body-part entities within one pair and one or more body part entities in a plurality of historical medical documents; clustering the non-body-part entities in the input medical document into one or more clusters based on the relevance scores of the all possible pairs; and outputting one or more body part entities related to one or more non-body-part entities clustered in each of the clusters as one diagnosis object.

Abstract: The present disclosure includes a method, apparatus and system for generating a neural network and a non-transitory computer readable storage medium storing instructions. The method comprises: recognizing at least an attribute of an object in a sample image according to a feature extracted from the sample image, using the neural network; determining a loss function value at least according to a margin value determined based on a semantic relationship between attributes, wherein the semantic relationship is obtained from a predefined table at least according to a real attribute and the recognized attribute of the object, wherein the predefined table is composed of the attributes and the semantic relationship between the attributes; updating a parameter in the neural network according to the determined loss function value. When using the neural network generated according the present disclosure, the accuracy of object attribute recognition can be improved.

Abstract: One of the aspects of the present invention discloses a feature point detection method. The method comprises: acquiring a face region in an input image; acquiring first positions of first feature points and second feature points according to a pre-generated first model; estimating second positions of the first feature points according to the first positions of the first feature points and pre-generated second models; detecting third positions of the first feature points and the second feature points according to the second positions of the first feature points, the first positions of the second feature points and pre-generated third models. According to the present invention, the final detected face shape could approach to the actual face shape much more.

Abstract: This invention provides a method for extracting a diagnosis object from a medical document comprises: extracting, from an input medical document, body part entities and at least one type of non-body-part entities and the relations between the body part entities and the non-body-part entities; obtaining, for each pair of all possible pairs of the non-body-part entities, a relevance score between two non-body-part entities within one pair, wherein the relevance score is obtained by using the relations between the two non-body-part entities within one pair and one or more body part entities in a plurality of historical medical documents; clustering the non-body-part entities in the input medical document into one or more clusters based on the relevance scores of the all possible pairs; and outputting one or more body part entities related to one or more non-body-part entities clustered in each of the clusters as one diagnosis object.

Abstract: The present disclosure provides a detection apparatus and method, and image processing apparatus and system. The detection apparatus extracts features from an image, detects objects in the image based on the extracted features; and detects key points of the detected objects based on the extracted features, the detected objects and a pre-obtained key point sets. According to the present disclosure, the whole detection speed can be ensured not to be influenced by the number of objects in the image to be detected while the objects and key points thereof are detected, so as to better meet the requirements of timeliness and practicability of the detection by the actual computer vision task.

Abstract: A detection method including extracting features from an image, detecting a human in the image based on the extracted features, detecting an object in a surrounding region of the detected human based on the extracted features and determining human-object interaction information in the image based on the extracted features, the detected human and the detected object. The detection speed and detection precision of detecting the human, object and human-object interaction relationship from the video/image can be enhanced, and therefore the timeliness and accuracy of offering help to the human in need of help can be better met.

Abstract: An image processing apparatus for extracting features from video frames of a video; and determining, for non-initial video frames, reference information of an object detected in a previous video frame thereof in a corresponding non-initial video frame with respect to object information of the object; and detecting an object from an initial video frame based on the features and detects an object from non-initial video frames based on the features and the determined reference information. The processing time of the object detection processing can be reduced, and the real-time requirements of object detection in the video can be better satisfied.

Abstract: A detection apparatus to extract features from an image; determine the number of candidate regions of the object in the image based on the extracted features, wherein the determined number of the candidate regions is decided by a position and shape of the candidate regions; and to detect the object from the image based on at least the extracted features and the determined number, position and shape of the candidate regions.

Abstract: A template determining apparatus including an attribute distribution determination unit configured to determine a distribution of a specific attribute in a plurality of images; and a template determination unit configured to adaptatively determine a template set from the plurality of images according to the determined distribution of the specific attribute of the plurality of images. Where the determined template set will be used for image normalization.

Abstract: An attribute recognition apparatus including a unit for extracting a first feature from an image by using a feature extraction neural network; a unit for recognizing a first attribute of an object in the image based on the first feature by using a first recognition neural network; a unit for determining a second recognition neural network from a plurality of second recognition neural network candidates based on the first attribute; and a unit for recognizing at least one second attribute of the object based on the first feature by using a second recognition neural network.

Abstract: Embodiments of the present disclosure provide a quantization method of feature vector, search method, apparatus and storage medium, wherein the quantization method comprises: setting quantization codewords; selecting, from the set quantization codewords, at least one quantization codeword of which distance from an original feature vector is smaller than first predetermined distance as a quantization codeword subset; determining quantization feature vectors corresponding to the original feature vector by using at least two quantization codewords, wherein at least one quantization codeword from the at least two quantization codewords are in the quantization codeword subset. The search method is a method of performing a search among quantization feature vectors after a quantization using the quantization method.

Abstract: This invention provides methods and apparatuses for segmenting text. A method for segmenting a text including a plurality of sentences comprises: extracting a plurality of evidences and a plurality of inferences from the text; for each of said inferences, determining a preferred position for each of said evidences based on the text and/or segmentation histories, wherein the preferred position represents the position that the evidence is the most likely to take in a sequence of evidences which are used to make the inference; and segmenting the text into a plurality of sections by determining one or more of boundaries between every two consecutive sentences in the text as section boundaries based on the preferred positions for the evidences. By virtue of the present invention, the segmentation will be more accurate.

Abstract: One or more embodiments relate to an object indexing method, object indexing device and object indexing system. An object indexing method may include an initializing step of dividing at least one first data container and at least one second data container, and initializing respectively the first and the second data containers by using a first data indexer and a second data indexer; a first indexing step of adding the input feature vector of a specific object into a first data container currently being filled by using the first data indexer; and a second indexing step, when the data size in the first data container being filled has reached a threshold, selecting another first data container as the first data container currently being filled, and in the case of executing the first indexing step, converting bulk data in the first data container previously being filled into one available second data container.