1. Configurable AI to Align with Your Decision Criteria You can configure this AI according to your own decision criteria. KrackNet™ allows you to set your own configuration setting value so that the analysis is performed based on your specific criteria. A list of crack codes, terminology and workflow are avaible below. [Workflow Diagram](/site-content/workflow) [Terminology](/site-content/terminology) [Distress Code List](/site-content/code-list) [Understanding Road Distress](/site-content/distress) 2. Optimizing Your Configuration Prepare max upto 20 sets of your road scan images. Upload these files using the default configuration settings and start 'Free Tuning Run' Download the analyzed images and JSON data, and review the outputs based on your decision criteria. Based on your review, you can adjust the threshold values for each crack code, sensitivity levels, and crack size. Upload the images again with the new configuration settings and repeat until you achieve the desired results. Once you have the optimal configuration, you can apply this final configuraiton for real Paid Analysis by saving the configuration task name for use in actual mass processing. Note: The entire Configuration setting optimization process is provided free of charge until you finalize the best configuration. 3. Starting Real Analysis We charge 1 credit per 1 m image analyzed. Calculate the number of images you want to analyze. Purchase credits through PayPal, and we will recharge your account accordingly. For the credit price, please refer to the [Pricing](/site-content/pricing). The number of credits will be deducted based on the total number of images processed/analyzed in ' 4. Downloading You can download the data to your desktop. When you extract the downloaded compressed file, you will find a SUMMARY.CSV file that summarizes the length or area of damage for each image, along with JSON files containing the analysis results for each image. You can read these JSON files and convert them into a format or insert to SQL DB suitable for your business. The following is an example of the Output JSON and an explanation of each key value. { "imagePath": "P21-0010-E1-2_400_s008370000.jpg", // Analyzed image name. This JSON file must be placed in same folder with image file "imageData": null, // Ignorable, used by LabelMe "imageHeight": 10000, // Image height "imageWidth": 3581, // Image width "pavement": { // Pavement information "material": "concrete", // Pavement material: concrete or asphalt "score": "1.00" // Material prediction confidence }, "date": "2024-03-22", // Anaysis date "model": "KrackNet 3.5", //KrackNet model version "summary": { // Summary of analysis "crack": 0, // Total length of cracks (m) "patching": 0, // Total area of pachings (㎡) "crazing": 0, // Total area of crazings (㎡) "spalling": 0, // Total area of spallings (㎡) "lane_lines": [ // x coordinate of both lane lines [left, right] 364, 3549 ], "version": "5.0.1", // Compatible version of LabelMe "flags": {}, // Ignorable, used by LabelMe "shapes": [ { "description": "Longitudinal Joint", // Distress description "score": 0.6791343331336975, // Distress prediction confidence "density": "M", // Condition level of distress (L/M/H: Light/Medium/Heavy) "pid": 2, // Distress # of this analysis "label": "CPF4LN", // KrackNet Code "points": [ // Point data of crack line or area, form of [x, y] coordination sequence [3358.0, 4700.0], [3058.0, 5300.0] ], "group_id": null, // Ignorable, used by LabelMe "shape_type": "linestrip", // Type of point data (rectangle is area box, linestrip is polyline or polygon) "flags": {} // lgnorable, used by LabelMe }, ] } 5. Reviewing Downloaded Data You may have your own image viewer or labeling tool, but if not, please download the public free LabelMe toolkit provided. [LabelMe.zip](https://github.com/labelmeai/labelme/releases/download/v5.4.1/Labelme.exe) to download.