Task: A unit of analysis that includes the followings List of files to analyze Physical files on the hard disk to be analyzed Task Configuration Analysis results (analysis log, JSON, Summary.xlsx) Task Configuration: A set of options for analyzing the physical length of images and various analysis options Settings: Individual setting within the Task Configuration Setting Value: Individual setting value within the Task Configuration Active Task: The single Task currently in progress (tuning/testing, Paid Analysis in progress, Paid Analysis completed). Only one Active Task can be executed on each computer. Additional tasks must be performed on a physically different computer. Turtle Crack Merge Factor: When cracks are densely formed, instead of treating each individual crack separately, a merge factor value is used to combine the cracks into a larger area of concentrated cracking. For example, if the merge factor value is set to 0.7, then if the crack density in a particular area exceeds 70%, the cracks will be merged into a crazing area. Additionally, if the crazing areas cover more than 70% of a particular region, they will be recursively merged into an even larger crazing area. This approach is necessary to handle areas where cracks are closely clustered, rather than treating each crack individually. Crack Detection Method: 1pass: Performs basic crack detection 2pass: Based on the crack information detected in 1pass, it redefines the detection area slightly off from the existing crack detection area, performs a secondary detection, and integrates the results with the 1pass outcome. The crack detection processing time may take longer, but the results can become more refined. Sensitivity: It has a value ranging from 0.0 to 1.0. The types of damage affected by the Sensitivity setting are Traversal & Longitudinal Cracks, Patching, and Spalling. It is only required for KrackNet sub-models that use machine learning semantic segmentation. Segmentation determines the presence of damage on a pixel-by-pixel basis, and sensitivity is the criterion for determining whether a pixel is valid or not. The higher the sensitivity value, the more pixels are considered valid. To detect fine and subtle damage, the value should be increased. Threshold: It also has a value ranging from 0.0 to 1.0. All types of damage are affected by the Threshold settings, as it acts as a decision boundary for determining whether to recognize an object as damage or not, based on the AI's confidence level. For models using segmentation, if the average value of connected or adjacent valid pixels exceeds the threshold, it is finally classified as damage; otherwise, it is classified as a normal area. For models using object detection, the classification is based on the confidence of the object. Unlike sensitivity, the higher the threshold value, the less damage is detected, and the lower the value, the more damage is detected. Conversely, the accuracy of detected damage tends to improve as the threshold value increases. To ignore damage with low confidence, the threshold value should be set higher.