Aiforia®️ Create is the most versatile tool for developing deep learning AI models for image analysis. Its cloud-based, collaborative working environment allows multiple users to work together in real time, anywhere in the world. Let’s explore the remote validation features in more detail.
The validation features give the end user an easy interface for defining validation sets and collecting validation annotations. They also provide a convenient way to invite colleagues or consultants to give blinded scoring or diagnoses according to intended use criteria. External moderates can be invited to collaborate without the need for them to be registered Create users.
The tool's cloud-based nature makes this collaboration even easier. You can invite validators with a simple click of a button directly from your image.
The validation features then provide a human versus AI comparison and human versus human cross-comparison, generating important metrics for quality assurance, regulatory, and publication documentation.
All cross-comparisons are performed at the pixel level for semantic segmentation, and the Jaccard index is utilized for object evaluation. Pixel-level operation is essential in thoroughly testing the validity of your AI models. Relying solely on simple correlation methods can overestimate your model's performance and potentially obscure significant errors.
Particularly in biology, it is crucial to not only determine the number of counts but also to accurately identify their exact nature. Failing to do so could lead to misleading results, such as a hypothetical scenario where two different types of equally distributed objects are counted from IHC. In this case, validation might show 100% concordance, even if the actual agreement between operators is 0%.
Get to know other Aiforia Create features: