Neural Video Decoding With Differential Network Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for decoding compressed video data using artificial neural networks are inefficient due to high latency and computational requirements for determining the optimal decoding network, and require excessive data transmission.
Innovation Solution
The method involves differential encoding of a decoding artificial neural network relative to a reference network, allowing for efficient determination and reduced data transmission by including differential encoding data in the data stream, along with flags to indicate potential changes in the decoding network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the optimal decoding artificial neural network is determined through conventional methods, then decoding accuracy is improved, but latency time and computational resources increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-defining multiple candidate decoding artificial neural networks with different architectures and characteristics before the actual decoding process. These candidate networks are prepared in advance and stored, allowing the system to quickly select from pre-prepared options rather than determining the optimal network during decoding, thus reducing latency time while maintaining decoding accuracy.
2Measurement precision
If the optimal decoding artificial neural network is determined through conventional methods, then decoding accuracy is improved, but computational resources are excessively consumed
Solution Approach 1:
The patent applies partial action by selecting and using only the specific candidate decoding artificial neural network that is most suitable for the current data characteristics, rather than evaluating or deploying all possible networks. The system partially utilizes the pre-defined candidate networks based on actual needs, reducing computational resource consumption while maintaining decoding accuracy through selective use of appropriate network architectures.
3Adaptability or versatility
If complete decoding artificial neural network data is transmitted in the data stream, then network adaptation is improved, but data transmission volume increases
Solution Approach 1:
The patent applies the extraction principle by separating the candidate decoding artificial neural network data from the main data stream transmission. Instead of transmitting complete network data, the system extracts and transmits only essential identification information or parameters that allow the receiver to identify and select the appropriate pre-defined candidate network. This significantly reduces data transmission volume while maintaining network adaptation capability.
4Adaptability or versatility
If multiple candidate decoding artificial neural networks are pre-defined, then adaptability to different data characteristics is improved, but device complexity increases
Solution Approach 1:
The patent applies the copying principle by creating multiple candidate decoding artificial neural networks that are copies or variations of a base network architecture. These candidate networks share common structural elements and parameters, allowing the system to maintain adaptability through multiple options while reducing overall complexity through architectural reuse and standardization. The candidate networks are copies with modified parameters rather than entirely different complex systems.
Data Source
AI summary
A part of a data stream includes a plurality of data units respectively associated with a plurality of images and together presenting different images of the plurality of images. The data stream part further includes differential encoding data differentially encoding an artificial neural decoding network relatively to a reference artificial neural decoding network. A method for decoding this data stream part includes the following steps: determining the artificial neural decoding network by decoding the differential coding data; and decoding at least one data unit of the plurality of data units by the determined artificial neural decoding network. Another decoding method, decoding devices, a computer program and associated data streams are also described.