Failure Bank · Quick start
Review model mistakes and decide which evidence should teach the next model or protect a regression.
Before you start
Open a project and choose available dataset images and model evidence.
Quick start
- Open Review, choose dataset, split and model. Import latest training evidence or run AI prediction when needed.
- Inspect the model prediction against human annotations. Describe the failure and choose its route.
Functions and choices
Explore the guide
Use numbered tabs and topic buttons to explore. Overview returns to the introduction; Manual opens help.
Choose a route
Training evidence, regression guards and uncertain cases have different roles.
Follow the walkthrough
Watch tutorial opens step 06, Improve with evidence, with an offline film and English captions.
Result
Reviewed failure evidence with preserved context.
Next step
Correct training-worthy annotations in Annotator and publish the next dataset version.