Abstract: Disclosed is a method for coding a video based on a long-term reference frame. The method includes: setting a long-term reference frame according to attribute information of an image frame; determining a reference index of a to-be-coded image frame based on a normal reference frame and the long-term reference frame; and acquiring a target matching block by performing inter prediction based on the reference index, so as to achieve coding of the to-be-coded image frame; wherein setting the long-term reference frame according to the attribute information of the image frame includes: setting the long-term reference frame based on a degree of redundancy in a time domain and a degree of redundancy in a spatial domain of the image frame.

Abstract: Provided is a method for moderating abnormal users. The method includes: acquiring history behavior data of a plurality of to-be-moderated users, wherein the history behavior data includes behavior data of interactions of the plurality of to-be-moderated users; extracting a plurality of predetermined valid features from the history behavior data for each of the plurality of to-be-moderated users, wherein the predetermined valid features are features predetermined in sample data; calculating, based on predetermined probabilities of events associated with the plurality of valid features, a probability that each of the plurality of to-be-moderated users is the abnormal user; establishing a total probability function based on probabilities that the plurality of to-be-moderated users are the abnormal users; determining candidate users by solving, based on predetermined conditions, a maximum value of the total probability function; and determining the abnormal users by moderating the candidate users.

Abstract: A method for detecting a melody of an audio signal, including: dividing the audio signal into a plurality of audio segments based on a beat, detecting a pitch frequency of each frame of audio sub-signal in each of the audio segments, and estimating a pitch value of each of the audio segments based on the pitch frequency; determining a pitch name corresponding to each of the audio segments based on a frequency range of the pitch value; acquiring a musical scale of the audio signal by estimating a tonality of the audio signal based on the pitch name of each of the audio segments; and determining a melody of the audio signal based on a frequency interval of the pitch value of each of the audio segments in the musical scale.

Abstract: A method for processing images includes: determining an image attribute of each image block of a plurality of image blocks in a to-be-processed image, determining a filter model corresponding to each of the image blocks based on the image attribute of each of the image blocks, and acquiring a filtered image of the to-be-processed image by filtering each of the image blocks by the filter model corresponding to each of the image blocks.

Abstract: Provided are a method for pushing anchor information, a computer device, and a storage medium. The method for pushing the anchor information includes: recalling an anchor user; determining an anchor type of the anchor user by using a live streaming history and a live streaming efficiency as classification dimensions; calculating an interaction score of the anchor user, wherein the interaction score represents a feature of a viewer user viewing a live streaming room of the anchor user; calculating a comprehensive score of the anchor user based on the interaction score and the anchor type; and pushing anchor information of the anchor user to the viewer user based on the comprehensive score.

Abstract: Provided is a method for recommending videos. The method includes: acquiring an identifier of a video operated by a target account in history, and determining a tag weight of each of a plurality of video tags of the video corresponding to the identifier; determining, based on the tag weight of the each video tag, a video weight of each of a plurality of candidate videos in a predetermined candidate set; selecting, based on the video weight of the each candidate video, a target video from the candidate set for recommendation; and detecting, from video tags of a distributed video, a video tag that meets a predetermined condition, and reducing, based on the tag weight of the video tag that meets the predetermined condition, a weight of an undistributed candidate video with the video tag that meets the predetermined condition in the candidate set.

Abstract: A method for transforming an audio signal comprises obtaining a plurality of segmental original frequency-domain signal segments and a plurality of segmental target frequency-domain signal segments by segmenting and performing a Fourier transform on an original audio signal and an initial target audio signal obtained by pitch shifting on the original audio signal; obtaining a plurality of original formant envelopes by respectively filtering the plurality of segmental original frequency-domain signal segments according to a plurality of original segment window functions, and obtaining a plurality of target formant envelopes by respectively filtering the plurality of segmental target frequency-domain signal segments according to a plurality of target segment window functions; and determining a pitch-shifted audio signal based on the plurality of segmental target frequency-domain signal segments, the plurality of original formant envelopes, and the plurality of target formant envelopes.

Abstract: A method for training a neural network, including: determining a neural network; training the neural network at a first learning rate according to a first optimization mode, where the first learning rate is updated each time the neural network is trained; mapping the first learning rate of the first optimization mode to a second learning rate of a second optimization mode in the same vector space; determining the second learning rate satisfies a preset update condition; and continuing to train the neural network at the second learning rate according to the second optimization mode.

Abstract: A method for processing images, including: acquiring an original image and auxiliary information of the original image; acquiring an object feature map by inputting the original image into a main path of a first visual task processing model, and acquiring an auxiliary feature map by inputting the auxiliary information into a branch path of the first visual task processing model; and acquiring a response map of the original image by fusing the object feature map and the auxiliary feature map and inputting fused object feature map and auxiliary feature map into the main path of the first visual task processing model.

Abstract: Provided is a message display method, including: displaying a message display interface of a multi-user communication group; displaying a message of a target user account in the multi-user communication group in the message display interface in a first display mode; displaying a shortcut move region; displaying the message of the target user account in the message display interface in a second display mode in response to a drag operation for dragging the account identifier of the target user account from the first account display region to a second account display region; and canceling displaying of the shortcut move region in response to not receiving an operation on the shortcut move region within a predetermined period.

Abstract: A backend architecture method includes: acquiring transaction information from a client, wherein the transaction information includes location information of the client; assigning, based on the location information, the transaction information to a transaction node that has jurisdiction over the location information, wherein the transaction node is a transaction center within a specified region in a distributed transaction system; matching, according to the transaction node, a channel gateway in a channel docking layer that pre-establishes a connection relationship with the transaction node, and sending a transaction request in the transaction information to a transaction object over the matched channel gateway such that the transaction object processes the transaction request; and acquiring a processing result from the transaction object, and transmitting information of the processing result to the client.

Abstract: Provided is a system for pushing information. The system for pushing information includes: a time-driven scheduling terminal, a push content generating terminal, and a push content candidate terminal; wherein the time-driven scheduling terminal is configured to calculate push time points of push users for different push service types according to push time point rules for the different push service types, and initiate content generation requests based on push service types and user bases at same push time points; and the push content generating terminal is configured to select, based on the content generation requests, target push content of the push users at the push time points corresponding to the push service types from the push content candidate terminal using a dynamically configured content push algorithm, and feed the target push content back to the time-driven scheduling terminal.

Abstract: A method for recognizing a video action includes determining an action category and action positioning information of a current video frame based on the current video frame and at least one forward video frame; and determining action content of a video based on the action category and the action positioning information of the current video frame.

Abstract: A method for recognizing actions in a video is provided. The method comprises, determining, based on key points of skeleton of at least one video frame in a video to be recognized, shallow features of the at least one video frame; acquiring image features of the at least one video frame; acquiring action features of the at least one video frame based on the shallow features and the image features; acquiring an action recognition result of the at least one video frame by inputting the action features of the at least one video frame into an action recognition model, wherein the action recognition result includes states and state probabilities of actions in the at least one video frame; and determining an execution state of a target action in the at least one video frame based on the action recognition result of the at least one video frame.

Abstract: Provided is a method for detecting line-of-sight. The method for detecting line-of-sight includes: determining, based on a key feature point in a face image, a face posture and an eye pupil rotational displacement corresponding to the face image, wherein the eye pupil rotational displacement is a displacement of a pupil center relative to an eyeball center in the face image; and acquiring a line-of-sight direction of an actual face by back-projecting, based on a preset projection function and the face posture, the eye pupil rotational displacement to a three-dimensional space where the actual face is located.

Abstract: Provided is a method for encoding video data. The method includes: acquiring video data; configuring an encoding parameter of an encoder; determining a first time interval between two adjacent video frames in a group of pictures based on the encoding parameter; and adjusting a number of video frames in the group of pictures based on the first time interval by encoding the video data based on the encoding parameter. A device and a storage medium are also provided.

Abstract: Provided is a method for moderating live streaming. The method includes: performing preliminary violation moderation on a current live streaming frame in a to-be-moderated live streaming room based on a high-accuracy moderation model and a high-recall moderation model that are cascaded with each other; and acquiring a target violation score of the current live streaming frame by inputting a first violation score of the current live streaming frame under the high-accuracy moderation model, a second violation score of the current live streaming frame under the high-recall moderation model, and a multi-dimensional behavior feature of the current live streaming frame in the to-be-moderated live streaming room to a pre-constructed behavior moderation model in response to the current live streaming frame passing the preliminary violation moderation.

Abstract: Provided is a method for reconstructing a face mesh model. The method includes: acquiring face scanning data to be reconstructed and a three-dimensional face mesh template; obtaining a target face mesh model by hierarchically extracting key feature points in the 3D face mesh template and by sequentially deforming the 3D face mesh template based on posture matching positions of the hierarchically extracted key feature points in the face scanning data; and obtaining a reconstructed face mesh model by acquiring global feature points in the target face mesh model and by deforming the target face mesh model based on the posture matching positions of the global feature points in the face scanning data.

Abstract: Provided is a video playback method, applicable to an anchor client. The method includes: displaying a video selection interface in response to a video display operation on a target account, wherein the video selection interface includes at least one video posted by the target account; sending a video playback request to a video server in response to a select operation on a target video in the video selection interface, wherein the video server is configured to transmit the target video to a live streaming server in response to the video playback request, and the live streaming server is configured to insert a video frame of the target video into a live streaming frame corresponding to the anchor client and push a live stream; and displaying the live streaming frame with the video frame inserted.