US Patent Application 18333067. SYSTEM AND METHOD FOR TRAINING ARTIFICIAL INTELLIGENCE MODELS FOR IN-LOOP FILTERS simplified abstract

From WikiPatents
Jump to navigation Jump to search

SYSTEM AND METHOD FOR TRAINING ARTIFICIAL INTELLIGENCE MODELS FOR IN-LOOP FILTERS

Organization Name

Samsung Electronics Co., Ltd.


Inventor(s)

Anubhav Singh of Bengaluru (IN)

Aviral Agrawal of Bengaluru (IN)

Raj Narayana Gadde of Bengaluru (IN)

H Keerthan Bhat of Bengaluru (IN)

Yinji Piao of Suwon-si (KR)

Minwoo Park of Suwon-si (KR)

Kwangpyo Choi of Suwon-si (KR)

SYSTEM AND METHOD FOR TRAINING ARTIFICIAL INTELLIGENCE MODELS FOR IN-LOOP FILTERS - A simplified explanation of the abstract

This abstract first appeared for US patent application 18333067 titled 'SYSTEM AND METHOD FOR TRAINING ARTIFICIAL INTELLIGENCE MODELS FOR IN-LOOP FILTERS

Simplified Explanation

The abstract describes a method for training AI models for in-loop filters in video codecs. Here is a simplified explanation of the abstract:

  • The method involves generating a training dataset by passing a video through a codec pipeline.
  • One or more predefined block features are extracted from the training dataset.
  • The extracted block features are used to create clusters, which group similar features together.
  • The clusters are further divided into sub-clusters based on the block features and a threshold for intra-cluster variation.
  • The sub-clusters are then supplied separately into multiple AI models.
  • The AI models are based on the extracted block features and are used for in-loop filtering in video codecs.

In summary, this method involves training AI models for in-loop filters by extracting block features from a video dataset, creating clusters, and dividing them into sub-clusters for improved performance in video codecs.


Original Abstract Submitted

An example method for training AI models for in-loop filters includes generating a training dataset by passing a video through a codec pipeline, extracting one or more predefined block features from the training dataset, creating a plurality of clusters based on the extracted one or more predefined block features from the training dataset, dividing the plurality of clusters into a sub-plurality of clusters based on the extracted one or more predefined block features and an intra-cluster variation threshold, and supplying the sub-plurality of clusters separately into a plurality of AI models based on the extracted one or more predefined block features.

Retrieved from "http://wikipatents.org/index.php?title=US_Patent_Application_18333067._SYSTEM_AND_METHOD_FOR_TRAINING_ARTIFICIAL_INTELLIGENCE_MODELS_FOR_IN-LOOP_FILTERS_simplified_abstract&oldid=6448"