From Cloud Storage to Trained Model: Mirror & Enrich Data from S3, Google Cloud, Azure
So you want to train a computer vision model. That means you need the right training data. You might already have the right data, but it isn’t well organized. It’s spread across years of images and video in cloud storage, and pulling the right examples into a project usually means a few engineers downloading files, writing one-off scripts, and guessing which folders hold what you need.
In this live session, Roboflow Full Stack Engineer Jeremy Prescott will show how to automatically mirror images and video from Amazon S3, Google Cloud Storage, or Azure Blob Storage into Roboflow, enrich them so you can search by what's in them, and use the right ones as training data, all without downloading a file. You'll also learn how to use the same setup to run the finished model across everything you store.
What you'll see:
Connect your bucket: Mirror images and video from Amazon S3, Google Cloud Storage, or Azure Blob Storage into Roboflow with read-only access, on a schedule that picks up new files.
Find the right data: Run a model over every file to add searchable metadata, then filter the Asset Library down to the images worth training on.
Label and train: Auto-label the selection with large AI models, then train a smaller model that runs fast in production.
Run it at scale: Point Batch Processing at the same bucket and run a Workflow with the new model across every file.