Lion Cipher Lab
Build a local image classifier
- AI Security
- Beginner
- ~35 min
- Free · no account required
Why it matters
Most people meet AI as something that talks, then hear that AI can now "see" and quietly assume it understands a picture the way a person does. It does not. A huge amount of practical computer vision is quieter and narrower than that: a model reads an image and sorts it into one of a fixed set of buckets, with a confidence number attached. Is this photo a cat or a dog? Is this X-ray flagged or clear? Is this uploaded image allowed or not? That is image classification, and it powers an enormous amount of real software precisely because it is small, fast, and cheap to run offline. Running one…
What you'll learn
- image classification
- Hugging Face transformers
- running a pretrained model offline
- reading labels and confidence scores
- recognizing out-of-distribution failures
- understanding model bias and limits
Steps in this lab
- What image classification actually is
- Install the tools, and how to run Python
- Classify one of your own images in three lines
- Read the top labels, not just the winner
- Run it across a few of your own images
- Push it off the map: confident and wrong
- The limits: fixed classes, bias, and false confidence
- What you can now do that most people can't