When it launched in early 2020, Roboflow was primarily a tool for converting object detection datasets. As the months have gone on we have expanded its capabilities to include collaboration, training, deployment, and annotation. We have supported uploading, augmenting, and converting classification datasets for models like OpenAI's CLIP but the core features of Roboflow were squarely focused on object detection.

Until now. Today we're announcing our intention to expand Roboflow's full capabilities to solve classification problems as well! We're starting with the addition of labeling images for Classification to Roboflow Annotate and will soon launch support for training and deploying classification models in Roboflow Train and Roboflow Deploy.

Choose "Classification" when creating a new project.

To get started, select "Classification" when creating a dataset. If you have images already grouped into folders, you can drop them into the upload widget.

Upload your labeled or unlabeled images.

If you don't already have labels for your images, upload your raw images and you can now label them inside of Roboflow with Roboflow Annotate.

Add or edit labels with Roboflow Annotate.

Try it today and stay tuned as we expand our support for new model types like Classification across the core product.