New project or existing project?
Whether you are adding more classes or just trying to improve performance on more types of data, it can be difficult to know whether to create a new project or add data to an existing project.
These are common scenarios when you almost always want to create a new project:
- Your imaging is dramatically different:
- Different magnification
- Different imaging modality (fluorescent vs. label-free)
- You want to add a region (not object) class that overlaps with a different region class (single projects only support object overlap
These are common scenarios when you want to create a new project:
- Very important to see all objects in one result
- Adding more classes of the same type (region/object) and want to optimize run credit cost
- You want object hierarchy (parent-child relationships) from your new class and an old one.
- You have lots of data and want the convenience of running images in one run.
In some scenarios, such as:
- Slightly different magnification
- Different tissue type or very different cell line
- Different coloring
- Different imager
In these scenarios, you'll likely get better performance with the same amount of labeled data if you create a new project, but you'll lose out of convenience of having it under one model.
Last modified 2mo ago