CV Vertex AI Trainer

CLI used to remotely train a Computer Vision model using the Google Vertex AI platform.

We are assuming:

• You have a Google Cloud Platform account
• You have a Google Cloud Platform project
• You have a Google Cloud Storage bucket
• You have a Label Studio project with images and annotations
• In order to retrieve your Label Studio token, refer to: https://api.labelstud.io/api-reference/introduction/getting-started#authentication
• Your images are stored in a Google Cloud Storage bucket

Setup requirements

1. Run pip install -r requirements.txt.

2. Install gcloud CLI, refer to: https://cloud.google.com/sdk/docs/install?hl=es-419

3. Run gcloud auth login.

   If you are getting a warning about "project quota" or about "default credentials not found" run gcloud auth application-default login.

4. Run pip install . to install the package (or pip install -e . to install in editable mode)

Usage

Create a .toml file following the example on config_example.toml

Run cv-vertex-ai-trainer -c config.toml to start training on the cloud.

Add --local to run the training locally.

Training with OBB

When training with OBB (Oriented Bounding Box) you need to uncomment the line in train_requirements.txt that installs Ultralytics from a forked repository. This is TEMPORARY until the original repository fully supports OBB.

Then you need to set the obb parameter to true in the configuration file, and pick a suitable YOLO model (for example yolov8n-obb.pt).