Visual Anomaly Detection Without Reference in Graphics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for detecting video corruptions rely on human intervention, which are prone to errors due to distraction and variability among engineers, and lack scalability as they require manual or reference-based methods that are inefficient for dynamic content like computer games and websites.
Innovation Solution
A novel automated no-reference anomaly detection system that uses deep learning techniques to identify visual corruptions in video streams without requiring a reference frame, enabling robust validation of graphics platforms and scalable testing across various graphics computing environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual or reference-based techniques are used for visual anomaly detection, then detection can be performed with existing tools, but human error and lack of scalability occur
Solution Approach 1:
The patent replaces manual human inspection with an automated deep learning-based anomaly detection system. The system uses neural networks to process video frames and identify visual corruptions automatically, eliminating human error and enabling scalable testing across multiple graphics platforms simultaneously.
Solution Approach 2:
The anomaly detection system is self-sufficient and does not require reference frames or pre-profiled expected outputs. The deep learning model independently identifies anomalies by learning normal visual patterns during training, allowing it to detect corruptions in dynamic content without external reference materials.
2Measurement precision
If reference-based methods are used for anomaly detection, then comparison with expected outputs is possible, but real-world dynamic content cannot be tested
Solution Approach 1:
The patent employs a dynamic anomaly detection approach where the deep learning model is trained on diverse video content to learn normal patterns. The system adapts to different types of content (games, websites, videos) and can detect anomalies in real-world dynamic content that continuously evolves, rather than being limited to static reference comparisons.
Solution Approach 2:
The anomaly detection system is designed to work universally across different types of video content including computer games, websites, and traditional videos. The deep learning model can handle various content types and anomaly types (visual corruptions, display issues, graphics errors) without requiring content-specific reference materials.
3Measurement precision
If human engineers perform visual anomaly detection, then detailed examination can be conducted, but time limits and distraction reduce efficiency
Solution Approach 1:
The patent replaces time-consuming manual inspection with automated deep learning-based detection. The system processes video frames rapidly using neural networks, identifying visual corruptions in seconds rather than requiring engineers to spend extended periods examining each video, thereby eliminating time limits and distraction-related errors.
Solution Approach 2:
The anomaly detection system operates continuously and automatically without interruption. The deep learning model can process multiple video streams simultaneously and continuously monitor for anomalies without the breaks, distractions, or fatigue that limit human engineers' working hours and efficiency.
Data Source
AI summary
A mechanism is described for facilitating visual anomaly detection without reference in computing environments. An apparatus of embodiments, as described herein, includes one or more processors to select a frame from a sequence of multiple frames associated with a video stream captured by a camera, and dynamically compute a frame confidence score for the frame based on frame training data associated with frame. The one or more processors are further to detect one or more anomalies in the frame when the frame confidence score is less than a frame confidence threshold associated with the frame, where detecting includes dynamically comparing the frame confidence score with the frame confidence threshold through inference using frame field data and the frame training data.