Hugging Face has expanded its AutoTrain platform to support computer vision, specifically image classification, allowing users to train models with zero coding required. AutoTrain simplifies machine learning by handling model selection and training automatically, making it accessible to non-coders and busy developers alike.
To get started, create a Hugging Face account and head to the AutoTrain homepage. Click "Create new project," name your project (e.g., butterflies-classification), select "Image Classification" as the task, and choose the "Automatic" model option. Then upload your dataset—either by dragging and dropping local files or by using a dataset from the Hugging Face Hub. You can optionally provide separate training and validation sets or let AutoTrain split the data.
Next, select the number of candidate models to try (training with 5 models and fewer than 500 images is free), review the expected cost, and start training. AutoTrain tests multiple architectures, prunes underperformers, and delivers the best model. For example, a butterfly classifier trained on the NimaBoscarino/butterflies dataset achieved 84% accuracy with minimal effort.
Once training completes, you can access your model on the Hugging Face Hub and test it using the integrated inference widget. AutoTrain's image classification feature opens up possibilities for recognizing signatures, identifying bird species, detecting plant diseases, and more.
Join the community on Discord at hf.co/join/discord for support and to share your projects.