A video frame prediction method based on multi-layer convolutional structure
By combining multi-layer convolutional structures and involution operators, an encoder, converter, and decoder were designed, solving the problem of high resource consumption in video frame prediction models and achieving efficient video frame prediction.
CN116567258BActive Publication Date: 2026-06-30UNIV OF ELECTRONICS SCI & TECH OF CHINA +1
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2023-05-18
- Publication Date
- 2026-06-30
Smart Images
Figure CN116567258B_ABST
Abstract
This invention discloses a video frame prediction method based on a multi-layer convolutional structure, belonging to the field of spatiotemporal data prediction technology using deep neural network models. The method has three main features: first, the proposed prediction model employs a multi-layer architecture; second, the convolutional structure of the prediction model uses a combination of multiple convolutional kernels of different sizes; and third, the prediction model uses an involution operator to replace the larger convolutional kernel. The proposed prediction model, while maintaining prediction accuracy, effectively reduces the inference time, computational load, number of model parameters, and memory usage of the video frame prediction task, significantly improving the efficiency of video frame prediction.
Need to check novelty before this filing date? Find Prior Art