Cognitive Motion Picture Analysis for Inconsistency Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Motion pictures often contain inconsistencies such as objects or sounds that are out of place in terms of time or location, which can detract from the viewing experience and are not effectively identified by existing technologies.
Innovation Solution
A computer-implemented method using machine learning trained models to analyze frames of a motion picture, identifying objects and their positions, and detecting inconsistencies such as changes in object position, appearance or disappearance, and incompatibility with the scene's time or location, providing information to adjust the frames accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to analyze frames to detect inconsistencies, then measurement precision of object positions and detection precision of inconsistencies are improved, but device complexity increases
Solution Approach 1:
The system segments the motion picture into individual frames and further segments each frame into multiple zones (first zone and second zone separated by horizon line). This segmentation enables targeted analysis of different regions with different requirements, improving measurement precision while managing complexity through divided processing.
Solution Approach 2:
The patent introduces an intermediary processing layer that includes horizon detection, zone segmentation, and object classification mechanisms. This intermediary layer acts as a mediator between raw frame data and inconsistency detection, breaking down the complex task into manageable steps that improve precision without overwhelming system complexity.
2Reliability
If frame analysis is performed to detect all types of inconsistencies, then reliability of motion picture quality assessment is improved, but loss of time increases
Solution Approach 1:
The analysis process is segmented into distinct stages: horizon line detection, zone division, object identification in each zone, and inconsistency checking. This segmentation allows the system to process frames systematically, improving reliability by ensuring comprehensive coverage while managing time through structured parallel processing of different zones.
Solution Approach 2:
The system performs partial analysis by focusing on specific zones (above and below horizon line) rather than treating all frames uniformly. By applying different analysis depths to different regions based on their importance, the system achieves high reliability for critical inconsistencies while reducing overall analysis time through selective processing.
3Manufacturing precision
If multiple zones are created for frame analysis, then manufacturing precision of inconsistency detection is improved, but device complexity increases
Solution Approach 1:
Each frame is divided into multiple zones based on the horizon line, with the first zone representing sky area and the second zone representing ground area. This segmentation improves detection precision by enabling zone-specific object validation (e.g., checking if objects appear in appropriate zones), while the automated horizon-based division keeps processing complexity manageable.
Solution Approach 2:
Different zones are assigned different quality requirements and analysis approaches. The first zone (sky) and second zone (ground) have different expected object types and inconsistency criteria. This local quality approach improves manufacturing precision by tailoring analysis to each zone's characteristics while simplifying complexity through context-specific processing rules.
Data Source
AI summary
A computer-implemented method, a system, and a computer program product are provided. Each frame of a scene of a motion picture is analyzed, using one or more machine learning trained models, to identify objects and respective positions of the identified objects in the scene. Inconsistencies in the frames are determined via at least one machine learning trained model, wherein the inconsistencies include one or more from a group of: a change of at least one position of the identified objects in the scene; an appearance or disappearance of at least one of the identified objects in the scene based on contiguous frames; and at least one of the identified objects being inconsistent with respect to a time period of the scene. Resulting information regarding the detected inconsistencies in the frames of the scene is provided to adjust the frames.