Video frame selection model training method, video processing method and device

The invention provides a training method of a video frame selection model and a video processing method and device. The training method comprises the following steps: acquiring a sample video; determining a target frame selection result corresponding to the sample video based on a to-be-trained video frame selection model; wherein the target frame selection result comprises a selected video frame, which is used for representing the characteristics of the sample video, in a plurality of candidate video frames contained in the sample video; determining a first image feature corresponding to the plurality of candidate video frames and a second image feature of the selected video frame based on an image feature extraction model; and training the to-be-trained video frame selection model based on the first image feature, the second image feature and the target frame selection result. 本公开提供了一种视频选帧模型的训练方法、视频处理方法及装置，包括：获取样本视频；基于待训练的视频选帧模型确定所述样本视频对应的目标选帧结果；其中，所述目标选帧结果包括所述样本视频所包含的多个候选视频帧中，用于表征所述样本视频的特征的选定视频帧；基于图像特征提取模型