README.md

metadata

title: Sign Identification Gui
emoji: 📈
colorFrom: purple
colorTo: gray
sdk: gradio
sdk_version: 5.47.2
app_file: app.py
pinned: false

🪧 Sign Identification (AutoML) — Gradio App

This app wraps a classmate’s image model cassieli226/sign-identification-automl (AutoGluon Multimodal).
Upload a sign image to see both the original and the preprocessed (224×224) image the model actually sees, plus class probabilities.

✨ What this app shows

• Image upload (PNG/JPG; webcam optional)
• Validation & friendly errors (type/size)
• Original vs. preprocessed image side-by-side
• Exposed inference parameter: Top-K
• ≥ 3 clickable Examples for quick testing

🔗 Model & Dataset

• Model: https://huggingface.co/cassieli226/sign-identification-automl
• Original dataset: ecopus/sign_identification
  • augmented split (385) for train/val
  • original split (35) for test

🛠️ Tech

• Framework: AutoGluon Multimodal
• Backbone: TimmAutoModelForImagePrediction (~194M params)
• Inference: predict_proba over 2 classes (0/1)

⚠️ Notes & Limitations

• Small, student-collected dataset (~420 images) → may not generalize.
• Binary labels only (0, 1).

🙏 Acknowledgments

• AutoGluon team
• Hugging Face Hub
• CMU 24-679 (Designing & Deploying AI/ML)

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference