Neural Network Input Validation for Computer Vision Algorithm Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional computer-controlled process systems rely on assumptions and constraints for algorithm development, leading to unpredictable or destructive outcomes when applied under irrelevant conditions, as they lack the ability to validate input data effectively, especially in complex tasks like image recognition.
Innovation Solution
A trained neural network is employed to classify input data as valid or invalid and recommend the appropriate algorithm for computer vision systems, using deep learning techniques to evaluate and select the best-fit algorithm based on input data characteristics, preventing further processing with invalid data and ensuring optimal outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a computer vision algorithm is programmed to control the application of a rust-proofing overcoat to a rectangular object, then the algorithm can successfully process rectangular objects, but it will fail or produce unpredictable results when presented with circular objects
Solution Approach 1:
The patent applies preliminary action by using a trained neural network to evaluate and classify input data before the main computer vision algorithm processes it. The neural network performs preliminary validation to determine if the input data is suitable for processing, preventing the main algorithm from executing on incompatible data types. This resolves the contradiction by ensuring reliability through pre-validation while maintaining adaptability through the neural network's ability to recognize various object types.
Solution Approach 2:
The patent introduces an intermediary component - a neural network - that acts as a mediator between the input data and the main computer vision algorithm. This intermediary evaluates the input data characteristics and determines whether the data is appropriate for the algorithm to process. The neural network translates diverse input formats into a validated format that the main algorithm can reliably process, thus resolving the contradiction between reliability and adaptability.
2Productivity
If a computer vision algorithm processes all input data without validation, then processing speed is maintained, but invalid or irrelevant input data causes unpredictable or destructive outcomes
Solution Approach 1:
The patent implements preliminary action by inserting a neural network validation step before the main processing algorithm executes. This preliminary evaluation quickly assesses whether input data is valid and suitable for processing, allowing the system to maintain high productivity by skipping validation for valid data while preventing unsafe processing of invalid data. The neural network's fast inference capability ensures that the validation overhead is minimal.
Solution Approach 2:
The patent extracts the validation function from the main processing algorithm and places it in a separate neural network component. This extraction allows the validation process to be performed independently and efficiently, separating the safety-checking function from the productivity-critical processing function. The neural network extracts and evaluates relevant features of the input data to determine validity without interfering with the main algorithm's processing speed.
3Device complexity
If traditional algorithms are used with fixed assumptions and constraints, then they are simple and easy to implement, but they cannot handle variations in input data characteristics
Solution Approach 1:
The patent introduces a neural network as an intermediary layer between the simple traditional algorithm and the variable input data. This intermediary handles the complexity of adapting to different data characteristics through learned patterns, while the traditional algorithm remains simple and unchanged. The neural network absorbs the adaptability requirements, allowing the rest of the system to maintain low complexity.
Solution Approach 2:
The patent segments the processing system into two distinct components: a neural network for handling data variability and adaptation, and a traditional algorithm for performing the core processing task. This segmentation allows each component to be optimized independently - the neural network for adaptability and the traditional algorithm for simplicity. The neural network segment evaluates input characteristics and prepares data in a format suitable for the traditional algorithm segment.
Data Source
AI summary
A methodology is presented for using neural network (NN) techniques to evaluate input data presented to a computer-controlled processing system. An initial evaluation is used to determine if the input data represents a valid product that is intended to be processed by one or more algorithms within the computer system. If the input data is determined to be invalid, the operation of the algorithm on the product is not initiated (or halted if previously started). Presuming a valid input is ascertained by the NN-based evaluation system, further classification and identifications may be performed to properly match the presented data with a particular system process, as well as select an optimum algorithm for preforming a given task from a set of possible algorithms that may be used for that task.