One-time purchaseOwn your eligible versionYour Mac. Your images.
A little heads-up. These are our planned launch prices. Purchasing and public downloads are not available yet. We will publish trial details when they are ready.
02 / Inside the workbenchOne license. The whole workflow.
Small details. Full picture.
From the first camera frame to the next model comparison. Here is what you can do with BadBob.
01
Capture Lab
Collect with a little more purpose.
Live camera capture
Choose a camera, capture images and work in a full-screen preview.
Three capture modes
Build class coverage, focus on a scenario, or capture freely without a guided sequence.
Model-assisted capture
Use a compatible model for live predictions. Pause AI independently of the camera.
Sessions & review
Revisit saved sessions, review your images and send selections to Annotator.
A place for every image
Route reviewed material to the training queue, a guard set, or save it for later.
02
Annotator
Turn images into useful ground truth.
Images, folders & lists
Import images or folders and organize selections into named image lists.
Boxes & contours
Draw bounding boxes and editable contours, assign project classes and refine the geometry.
AI suggestions
Analyze selected images with a compatible model, then review and edit its suggestions.
An editor that keeps up
Pan, zoom, fit the image and undo edits while you work.
Saved revisions
Save human annotations as revisions and revisit their history. Choose what goes forward to a dataset.
03
Data Studio
Know exactly what went into a model.
Versioned datasets
Publish manifests that reference specific images and saved annotation revisions.
Train, validation & test
Keep split membership explicit, with reserved validation and optional test images.
Import & export
Review imported data and class mappings, and prepare exports for the next part of your workflow.
Inspect the evidence
Check image and annotation references before preparing a training workspace.
Segmentation preparation
Derive masks from saved human contours and inspect the image, overlay and mask.
04
Training
Give your own images something to teach.
Create ML object detection
Train locally with Full Network (YOLOv2) or Transfer Learning (Apple Object Print).
Control the run
Choose a published dataset, algorithm, iteration budget and output location. Follow progress or stop a run.
Augmentation & preview
Preview brightness and contrast changes. Rotation and mirroring are explicit opt-ins; augmentation starts off.
Protected originals & splits
Augmentation applies to training images. Originals, validation, test and guard images stay unchanged.
Run history & model lineage
Keep configuration and dataset references with saved results. Register Core ML candidates for later use.
ExperimentalInstance segmentation is experimental. The supported training environment requires Apple Silicon, a separate local setup and contour masks. It produces a checkpoint rather than a Core ML detector.
05
Evaluation & comparison
Find out what actually improved.
Image evaluation
Evaluate compatible saved models on held-out data and inspect individual image results.
Reference & candidates
Compare compatible models on the same dataset split and evaluation protocol.
Saved comparisons
Revisit comparison results and inspect differences by class.
Guard sets
Keep reviewed regression examples separate and use them to check for unwanted changes.
Segmentation evaluation
Inspect masks and compare compatible segmentation results through the experimental segmentation path.
06
Failure Bank
Make the mistakes useful.
Prediction review
Inspect model predictions alongside saved human annotations.
Training evidence
Import evidence from a training run or run predictions with a selected compatible model.
Search & filter
Find relevant images and reviews without working through the whole collection.
Review & route
Save a review and route useful examples into the next round of annotation or dataset work.
An explicit improvement loop
You decide what becomes training data and when to train again. Reviews do not silently retrain a model.
07
Projects & everyday tools
Your work, with a place to belong.
Your own classes
Configure projects around the objects you want to recognize.
Local or shared storage
Work with project packages and image banks on local disks or shared storage.
Model catalog
Keep track of model references and reconnect files when their storage location changes.
Model Journal
Link models to dataset versions and keep provenance, notes and tags with their history.
Camera & environment settings
Set camera and inference defaults, choose storage roots and prepare training dependencies.
Help where you need it
Open the contextual manual, follow quick starts, search for a topic and bookmark useful pages.
Language choices
English, Swedish, German, French, Spanish and more. Choose your app language; missing translations fall back to English.
03 / Before you beginBring a Mac. Bring your curiosity.
Made for Mac. Built around you.
macOS 13 or later. Camera capture needs an available camera and permission to use it. Your project, image bank and output location need enough writable storage.
Training runs on your hardware. Dataset size, memory and the selected training path affect what you can train and how long it takes. Shared storage is supported; local storage can reduce file-transfer overhead.
Apple Silicon for segmentation. Create ML powers object detection. The supported experimental segmentation training environment requires an Apple Silicon Mac, a separate local setup and prepared masks. Setup may download tools or models.
Cloud compute & third-party subscriptions are not included.
04 / A few good questionsCuriosity is encouraged.
Good question.
01Is this a subscription?+
No. BadBob is a one-time purchase. The license is permanent for the version you buy and the updates included in your update period. There is no automatic renewal.
02What happens after the first 12 months?+
Your eligible version keeps working. If you want another 12 months of updates, the optional update purchase is €49 including VAT. You do not have to buy it to keep using your existing version.
03How many Macs can I use?+
The planned single-user license includes 2 Mac activations. The update purchase extends updates for an existing BadBob license; it is not a second standalone license.
04Can I buy or download BadBob today?+
Not yet. Public downloads and checkout are being prepared. The prices here describe the planned launch offer. Trial details will be published when the trial is ready.
05Does the price include cloud training or AI credits?+
Training uses your own Mac and storage. Cloud compute and third-party service subscriptions are not included. Preparing a training environment may require downloading tools or models.
06Is segmentation included?+
Yes, as an experimental workflow with contour annotation, mask preparation, training, evaluation and compatible model inference. The supported segmentation training environment requires an Apple Silicon Mac and a separate local setup. Its checkpoints are not interchangeable with Create ML object detectors.