Tencent Technology (shenzhen) Company Limited

Shenzhen Create Future International Trading Company

Ping an Technology (shenzhen) Co., Ltd.

Hui Guo

Jing Jing Xiao

Jianzong Wang

Jicheng Song

Jing Xiao

A method for training a retrieval model is performed by a computer device. The retrieval model includes an embedding vector network and a quantization index network. The method includes: obtaining n sample triplets; inputting basic feature vectors of the n sample triplets to the embedding vector network to identify, according to errors of feature vectors output by the embedding vector network, a first sample triplet set used for training the quantization index network; inputting the basic feature vectors of the n sample triplets to the quantization index network to identify, according to errors of quantization indexes output by the quantization index network, a second sample triplet set used for training the embedding vector network; training the quantization index network on the basis of the first sample triplet set; and training the embedding vector network on the basis of the second sample triplet set.

Method and apparatus for training retrieval model, retrieval method and apparatus, device and medium

Blanket

A method for video loop recognition includes determining a first encoding feature and a second encoding from a first video clip pair of a video. The first encoding feature is associated with first modal information, and the second encoding feature is associated with second modal information that is different from the first modal information. A network model that includes a first sequence model associated with the first modal information and a second sequence model associated with the second modal information is acquired. The method includes inputting the first encoding feature to the first sequence model that outputs a first similarity result, inputting the second encoding feature to the second sequence model that outputs a second similarity result, and obtaining a loop comparison result based on a comparison of the first similarity result with the second similarity result. The loop comparison result indicates a video type of the video.

Video loop recognition

This application provides a video retrieval method performed by a computer device. The method includes: performing feature extraction on an image feature of a query video to obtain a first quantization feature, obtaining a second candidate video with a high category similarity to the query video based on the first quantization feature, and finally taking a second candidate video with a high content similarity to the query video as a target video. The quantization control parameters are adjusted according to the texture feature loss value corresponding to each training sample to cause the target quantization processing sub-model to learn the ranking ability of the target texture feature sub-model, to ensure that the ranking effect of two sub-models tend to be consistent, and an end-to-end model architecture enables the target quantization processing sub-model to obtain the corresponding quantization feature based on the image feature.

Video retrieval method and apparatus, device, and storage medium

This application discloses a method for recognizing a video clip performed by a computer device. The method includes: determining a plurality of video frame pairs between a first video and at least one second video, each video frame pair comprising a first video frame from the first video and a second video frame from the at least second video with a similarity satisfying similarity conditions; fusing the first video frames in the plurality of video frame pairs on the basis of occurrence time differences of each video frame pair to obtain at least one candidate video clip in the first video; and determining at least one target video clip in the first video on the basis of the at least one candidate video clip, the at least one target video clip being within a target time range of the first video.

Method and apparatus for recognizing video clip, device, and storage medium

Water ball

Embodiments of the disclosure provide a data processing method and apparatus, a device, and a medium. The method includes: performing video analysis on video data of a target video to obtain a plurality of video segments; determining a video template associated with a target user from a video template database based on a user portrait of the target user, and obtaining at least one predetermined template segment and a template tag sequence in the video template; screening at least one video segment matching the template attribute tag of the at least one template segment; splicing the at least one matched video segment according to a position of a template attribute tag of each template segment in the template tag sequence as a video material segment of the target video; and pushing the video data and the video material segment to an application client corresponding to the target user.

Growing community of inventors

Shenzhen, China

Hui Guo