SHELF PIPELINE

Annotate, sort, and export shelf datasets.

Create a project, run the labeling pipeline, then curate crops into product bins.

Deployment Loading...
Checking auth...

Datasets

Captured by image-api. Pick batches, choose a split, and start a project.

Create Project

Spin up a new project from a dataset zip and detector models.

Drop dataset zip here or click to choose a file

Upload a zip containing a train and/or test folder. Layout.csv and reference_images are optional; if missing, use custom bins during sorting. Paired requires both splits; Train only and Test only require just the named one.

Show dataset mode info

Auto detect picks the right mode from the zip contents. Paired uses both train and test. Train only uses just train. Test only is for held-out evaluation — assemble produces only a test set and export skips train/val split.

Show dataset directory info

Dataset root (contains raw/train/images and optional raw/test/images, plus optional catalog).

Show shelf key info

Select a shelf from Postgres to auto-generate the sorting planogram bins. Example: 870504_41_4O_4.

Show shelf detector info
Show product detector info

Projects

Open a project to manage its runs.