I’m a beginner to labelbox (and also relatively new to using python…). I have a labelled data set in labelbox, and I would like to use the labelled data set to train a model to label input images (and then export this model for use in python). I apologise for the long question but what are the steps for me to do so? Any help would be appreciated! Thanks!
Given you have labeled data, you just need to create a model, a model run (iteration) and allocate the linked ontology, data rows (asset) and then labels.
import labelbox as lb API_KEY = "<API_KEY>" client = lb.Client(api_key=API_KEY) #retrieve the ontology you need to link ontology = client.get_ontology("<ONTOLOGY_UID>") # create Model (name has to be unique) model = client.create_model(name="My_first_Labelbox_Model", ontology_id=ontology.uid) # create Model Run (has to be unique) model_run = model.create_model_run("iteration 1") #link the data rows, here I will source them from a project project = client.get_project("<PROJECT_UID>") #retrieve the data rows from a your project data_rows = [dr for b in project.batches() for dr in b.export_data_rows()] #allocate the data rows to your model run allocate_data_rows = model_run.upsert_data_rows(data_rows) #add the label to your model run model_label_import = model_run.upsert_labels(project_id="<PROJECT_UID>") #Your model prediction (...) #import your prediction to your model run (name has to be unique per import) upload_job_prediction = model_run.add_predictions( name="My_first_prediction_import", predictions=predictions, )
- API_KEY can be obtained from : Create an API key
- All other ids can be retrieved from Labelbox directly : Customer support
- Prediction (predictions) format (depending on data type) are documented here : https://github.com/Labelbox/labelbox-python/tree/b6ded55eaf313bd4c995cc86277a6290be10766c/examples/prediction_upload
Finally I encourage you to use our documentation for further details : https://docs.labelbox.com/
Our SDK documentation can be found here : Labelbox Python API reference — Python SDK reference 3.49.1 documentation