Auto-labeling for Image Classification
1. What is Auto-labeling for Image Classification?
Manual labeling and autolabeling are available for single image classification.
Image data with labels (folders categorized compressed by class) and data without labels are all available for labeling AI.
1. Include labeling data in compressed files: Load data → Create customai → Add data (data before labeling) → Auto-labeling
2. Without labeling data in compressed file: Load data → Manual labeling → Create customai → Auto labeling
For data that does not contain the results, you can proceed with autolabeling by generating customai after proceeding with a small amount of manual labeling.
For data containing the results, auto-labeling can proceed immediately after creating customai and adding unlabeled data.
2. Create Auto-labeling for Image Classification
[NOTE] In this case, describe how to creat project without labeling.
1) Upload data
1-1) Click LABELING AI's Start Label button to go to DS2 DATASET.
1-2) Compress the image file and upload it as a zip file. Image files support png, jpg, jpeg, and gif.
1-3) "Data Settings" sets whether images in extrusion are labeled or not. If the uploaded file is a compressed file that classifies the image as a class-specific folder, check the option "Include labeling data within compressed files" and, conversely, click OK without the option check.
2) Create project
2-1) After selecting a project name, description, and category, click Next to complete the project creation.
2-2) Click the Create Project button to go to the Project Management window you created, and click "Labeling AI | Labeling" to confirm that the project list has been updated.
3) Add class
3-1) If the dashboard "Classification" is empty, you must add the class. Click "More" in the Classification or "Class" in the middle.
3-2) Sets the class name and color. Class names must be written in English, and only some special characters are available. After completion, click the Done button.
3-3) You are able to see the class names.
4) Manual labeling
You can create Custom AI by doing more than 10 manual labeling for each class, and a total of 10 manual labeling for continuous value data.
4-1) When class creation for 'Classification' is complete, click the Start Manual Labeling button to the right of the project name to start manual labelling.
4-2) Select the class that matches the data information at the bottom of the screen and click the Save and Next buttons to proceed with labeling.
If you enter the wrong labeling, you can click the Previous button to correct and save the data labels that have already been worked on.
5) Create Custom AI
5-1) When you have more than 10 objects per class, click the Create CUSTOM AI button to create artificial intelligence for auto-labeling.
5-2) Select the class you want to label and click Start AI Development.
5-3) Pop-up of AI development start appears. The beginning and end of modeling learning can be found through email and notification history.
6) Start Auto-labeling
6-1) Add the data you want to auto-labeled. (If unlabeled data has already been uploaded, you can auto-labeled without any additional data.)
6-2) Click the Start Autolabeling button to start autolabeling.
6-3) Starting and ending auto-labeling can be found in your email and notification history.
2-5. Advancement of inspection and labeling AI
1) You can verify that auto-labeled data is properly labeled, and correct mislabeled data.
2) Also, by regenerating CUSTOM AI with modified label data, you can upgrade it to a more accurate autolabeling model.