Pre-annotation tool



With Picsell.ia you can use several automation.

1 automation is built in -> DEXTR

But you can also configure automatic labeling with pre-trained model.

By default, you have access to a MASK-RCNN trained on COCO which gives good results for the classes present in COCO Dataset.

If you attached different models to your project, you will be able to select which model you want to use to perform labeling automation.

But to make sure that our models will detect only relevant objects for your use-case, you need to bind the model's predictions with your classes.

To do so, go to Projects -> your project -> settings -> label config and bind the predictions.

For each labels declared in your project, you can attach a Models prediction by selecting it from the scroll-down menu.