Creating a New Project
There are 3 ways to create a new project:
- Creating a new project from scratch
- Creating a new project from existing projects on the AIEX platform and finally
- Creating a new project by importing projects from an external location.
1. Creating a New Project from Scratch
To create a new project from scratch, simply click the “CREATE NEW PROJECT” button in your organization’s workspace. A modal will appear where you can enter the name, a number of tags to better classify your project, a description to help define the scope of the project and its features, for your project, as well as select the project type. While the name, tags, and description are optional and can be changed later, the project type is required and cannot be altered once the project is created. If you want to convert your project to a different type, you can do so by creating a new project. However, please be aware that this will not transfer any trained models or versions from the original project, and you will need to start training a new model from scratch.
2. Creating a New Project from Existing Projects on the AIEX Platform
You can use existing projects as a foundation to start a new project in AIEX. This can be useful if you want to reuse data from a previous project, or if you want to combine and generate new data from multiple projects. There are two ways to create a new project based on existing projects in AIEX:
- Duplicating a Dataset from an Existing Project
To duplicate an existing project in AIEX, simply click on the Icon next to the project name in your organization workspace and select “Duplicate Project.” A modal will open, asking you to name the newly duplicated project.
- Keep in mind that the process of duplicating a project may take a while, depending on the amount of data in the original project. If the number of images in the duplicate project does not match the original, try refreshing the page after a few moments. Please be patient as the duplicate project is being created.
When you duplicate a project:
- Merging Two or More Datasets from Existing Projects
To merge multiple datasets into a single new project in AIEX, click on the “CREATE A NEW PROJECT” button in your organization workspace and select “Merge Datasets.” Select the projects that you want to merge, and click “MERGE SELECTION.” In the subsequent modal, enter a name for the new project and select the appropriate project type. Click “Create Project” to create a new project, which will contain the merged data from the original projects. Please note that the original projects will remain unchanged and a new project, containing data copied from the selected projects, will be created.
Generally, when merging two or more projects:
It is possible to merge projects of different types in AIEX, such as Classification, Object Detection, and Segmentation. While some project types are compatible with each other for merging, others may not be. Let’s see what when combining projects of different types:
Segmentation -> Object Detection:
Segmentation projects are fully compatible with Object Detection projects. Converting a segmentation project into an object detection project will transform all the polygons into bounding boxes.
Segmentation -> Classification:
When attempting to convert a segmentation project into a classification project, only the images will be carried over to the new project.
Object Detection -> Segmentation:
Object detection -> Classification:
Object detection projects are not compatible with classification projects. When attempting to convert an object detection project into a classification project, only the images will be carried over to the new project. In short:
- Object detection projects are fully compatible with segmentation projects. After converting an object detection project into a segmentation project, all the bounding boxes are transformed into box-shaped polygons. You can add points to the generated polygons at your will. Please note that training a segmentation algorithm on boxes is not recommended and it is recommended to identify objects using polygons first.
- When converting a project into an incompatible project, all the images, regardless of their type, will carry over to the new project, though
- Classification projects are not compatible with other project types. Meaning that while the images carry over, the labels cannot be transformed into polygons or boxes. Furthermore, polygons and boxes cannot be transformed into labels as an image might contain several annotations from different categories.
button.
