Just a few years ago, businesses were scrambling to hire any and all machine learning talent, and at high cost. Happily, today, you might not need to hire any machine learning talent. If you’re using Scale to annotate your data, and you’re working to solve a computer vision problem, model training and deployment just became a lot easier with the new “Models” feature in Scale Rapid. Models in Rapid automatically trains state-of-the-art machine learning models for your business case and delivers them with a simple API call or download.
Rather than doing the heavy lifting of setting up a pipeline for model training, transfer learning, hyperparameter tuning and model hosting, the new Models feature in Scale Rapid enables all software engineers to train and deploy a model endpoint (accessible via API) in just a few hours. And if your embedded or edge use case requires a model on-device you can also opt to download the model we trained for you.
Here’s how models in Rapid work:
First, you upload your image data. Just as if you were kicking off a labeling project, you specify the exact goals for the model you intend to build. It’s essential to know whether you want to build a classification, segmentation, or object detection model, and what exactly you want to predict in images your future model has never seen before.
Or, if you’re already a Rapid customer, you’ll have already uploaded data, and very likely it already has high-quality labels. Congratulations: you’re just one click away from a live API model endpoint. We automatically train a model for classification, segmentation, or object detection (depending on your selection when you upload data), so you won’t have to worry about model selection, training, or hyperparameter tuning.
Once the model is trained, you can deploy it either via our automatically hosted endpoint or download the raw model file. Customers use the models trained in Rapid both directly for their business applications or to generate pre-labels in order to speed up the labeling process.
If you’re ready to enable your software engineers with ML, or unburden your ML team with the ongoing maintenance of models in production, check out models in Rapid today.