AI-Assisted Annotation

## AI-Assisted Annotation

## Module Principle
Utilize the image processing functions and algorithms of VG3 software combined with object detection to provide batch-assisted annotation functions for AI training tools. Functions include: Full Annotation, Incremental Annotation, and Add Dataset.

## Usage Scenarios
When annotation is required for model training, the large volume of annotations can lead to low efficiency. Using assisted annotation can improve annotation efficiency.

## Usage Method
Add Dataset: Directly create an annotation project.
Full Annotation and Incremental Annotation:必须先 use the AI training tool to train a small model (approximately 50 images). Then, use the recognition capability of this small model to batch annotate the required feature information.

### Pre and Post Modules
When using AI-Assisted Annotation normally, if you choose Full Annotation or Incremental Annotation, a pre-module is required. If the annotation boxes are horizontal rectangles, the pre-module is Object Detection. If the annotation boxes are rotated rectangles, the pre-module is Rotated Object Detection.

### Usage Example
[Record operation video]

## Parameter Configuration Description

![](/media/202601/assistTag_1768357211.png)
Parameter Description:
1. Assistance Mode:
    (1) Full Assisted Annotation: Create a new annotation project, automatically creating the dataset and annotation set.
    (2) Incremental Assisted Annotation: Select an existing annotation project and annotation set to add annotations in batches incrementally.
    (3) Add Dataset: Create a new annotation project, only adding a dataset and populating the dataset with data.
2. Detection ROI Type: Selection of the annotation rectangle type, horizontal or rotated.
3. Select Training Project: Enabled only in Incremental Assisted Annotation mode, retrieves the list of currently available training projects.
4. Select Annotation Set: Enabled only in Incremental Assisted Annotation mode. After selecting a training project, it automatically retrieves the corresponding dataset list.
5. Create Annotation Project: Enabled in Full Assisted Annotation and Add Dataset modes. Clicking Create will automatically generate an annotation project.
6. Image Quality: The quality of images saved to the dataset, typically used to control image file size.
7. Pause Assistance Function: Saving of images and annotation information starts only after this option is checked.

![](/media/202601/assistTagPort_1768357829.png)
Port Description:
Input Image: Refers to the source of images that need to be saved to the dataset.
Target Label: Enable subscription during assisted annotation. The data source is the target label from Object Detection or Rotated Object Detection. The type depends on the selected ROI detection type.
Rectangle Data: Enable subscription during assisted annotation. The data source is the rectangle array from Object Detection or Rotated Object Detection. The type depends on the selected ROI detection type.

## Output Description
Only displays the execution result and error messages.
Click here to jump to the corresponding Chinese manual article:&nbsp;[AI-Assisted Annotation](/doc/168)