Demo
Draw one of the supported shapes and let the model classify it.
- Draw on the canvas (circle, triangle, square, hexagon, or octagon).
- Click “Classify” to submit the drawing.
- Use the eraser button (or press C) to clear and try again.
- Review the confidence bar and per-shape scores on the right.
STAR Summary
- Situation
- I wanted a model that can tell what shape someone drew.
- Task
- Owned the end-to-end build, from implementation through the final deliverable.
- Action
-
- Used Google's QuickDraw sketches and built train/validation splits.
- Trained a small ResNet18 classifier in PyTorch.
- Deployed a CPU-only AWS Lambda endpoint so the browser can request predictions.
- Result
-
- High accuracy on a five-shape subset (results vary by split and drawing style): circle, triangle, square, hexagon, and octagon.
- Predictions return quickly once the Lambda container is warm; cold starts can take longer.
Links
Data Links
Notes
The live demo runs CPU-only serverless inference on five QuickDraw shape classes, so confidence is scoped to that small drawing task.
