Combining Human and Machine Capabilities for Improved Accuracy and Speed in Visual Recognition Tasks
In various areas of pattern recognition, in particular, in visual recognition tasks, there is always a trade-off between accuracy and speed. Performing the task manually by an expert typically results in high accuracy, but takes a significant amount of time to accomplish. The visual recognition task areas can be facial recognition, flower recognition, hand-writing recognition, etc. Manual activity will involve comparing items within a known and presumably correct database. Performing the task automatically by machine typically can be done in a split second, but the accuracy level is quite low. ^ With these known conditions, there is room to investigate what level of accuracy and speed that can be achieved if we combine human and machine functionalities in visual recognition tasks. Human and machine collaboration can range in a broad spectrum, from utilizing a tool to assist an expert to utilizing humans in improving automated tools. In this research we will focus on improving the automated feature extraction process by adding limited human interaction where the time utilized is still acceptable. Thus, the problem statement here is: ^ “To investigate human assistance in an automated feature extraction of visual pattern recognition where higher accuracy is required than is currently achievable by automated systems through exploring various possible human assistances within a reasonable time.” ^ It is expected that an automated shape/contour recognition process with human assistance in color recognition will provide the highest accuracy while still maintaining an acceptable time in the process. This model can be used to improve existing visual pattern recognition tools or to create new ones ^
Information technology|Information science|Computer science
"Combining Human and Machine Capabilities for Improved Accuracy and Speed in Visual Recognition Tasks"
(January 1, 2016).
ETD Collection for Pace University.