Update README.md

README.md

@@ -10,37 +10,42 @@ license: mit
 ---
 # Book Genre Predictor
 
-This Space hosts a **Gradio app** that predicts the **numeric genre code of a book** based on its **physical dimensions and page count**.  
-It was built using a **tabular AutoGluon model** and deployed on Hugging Face Spaces.
+This Hugging Face Space hosts a **Gradio app** that predicts the **genre of a book** based on its **physical dimensions and page count**.  
+It uses a **AutoGluon Tabular model** trained during last session.  
 
 ---
 
-## Dataset & Model
+## Dataset & Model Card
 
-- **Source Model Repo:** [FaiyazAzam/24679-tabular-autolguon-predictor](https://huggingface.co/FaiyazAzam/24679-tabular-autolguon-predictor)  
-- **Task:** Predict `Genre` of a book given its physical features.  
-- **Features Used:**  
-  - `Height` (cm)  
-  - `Width` (cm)  
-  - `Depth` (cm, spine thickness)  
-  - `Page Count` (integer)  
+- **Dataset:** Book metadata dataset (features: `Height`, `Width`, `Depth`, `Page Count`; label: `Genre`).  
+- **Model Repo:** [FaiyazAzam/24679-tabular-autolguon-predictor](https://huggingface.co/FaiyazAzam/24679-tabular-autolguon-predictor)  
+- **Framework:** [AutoGluon Tabular](https://auto.gluon.ai/stable/index.html)  
+- **Task:** Multi class classification -> predict `Genre` (numeric code).  
 
-The model was trained using [AutoGluon Tabular](https://auto.gluon.ai/stable/index.html).  
-Prediction outputs are **numeric labels** (e.g., 0, 1, 2) that correspond to genres in the training data.
+### Input Features
+| Feature      | Type    | Unit / Description                 |
+|--------------|---------|-------------------------------------|
+| Height       | float   | cm – height of the book             |
+| Width        | float   | cm – width of the book              |
+| Depth        | float   | cm – spine thickness                |
+| Page Count   | integer | number of pages                     |
 
----
+### Label
+- `Genre` → encoded as **numeric codes** (e.g. 0, 1, 2, …).  
+- Mapping to actual names was not provided in the original dataset. 
 
-## App Instructions
+---
 
-1. Enter values for **Height, Width, Depth, Page Count**.  
-2. Click **Predict** to see the model’s prediction.  
-3. Use one of the **example inputs** to quickly test the app.  
+## App Interface
 
-✔️ Input validation ensures all values must be **positive numbers**.  
+- **Widgets:** Numeric input boxes for each feature.  
+- **Output:** Numeric code prediction (e.g. `"Predicted Genre: 1"`).  
+- **Examples:** 3 preloaded examples for quick testing.  
+- **Validation:** Ensures all inputs are positive.  
 
 ---
 
-## Example Inputs
+## 🔍 Example Usage
 
 | Height (cm) | Width (cm) | Depth (cm) | Page Count |
 |-------------|------------|------------|------------|
@@ -48,24 +53,45 @@ Prediction outputs are **numeric labels** (e.g., 0, 1, 2) that correspond to gen
 | 24.0        | 15.0       | 2.2        | 320        |
 | 18.5        | 12.0       | 1.5        | 180        |
 
+Note: The model often defaults to predicting a single genre (e.g. Fiction / code 0).  
+This reflects dataset/model limitations, not the app itself.
+
 ---
 
-## Technical Notes
+## Technical Details
 
-- **Framework:** [Gradio](https://www.gradio.app/) interface.  
-- **Backend:** AutoGluon `TabularPredictor`.  
+- **Backend:** AutoGluon `TabularPredictor` loaded from a zipped artifact.  
+- **Interface:** [Gradio](https://www.gradio.app/).  
 - **Deployment:** Hugging Face Spaces (`sdk: gradio`).  
-- **Known Limitation:** Output is a **numeric genre code**, since the training dataset only contained encoded labels.  
+- **Environment:** Python 3.10, pinned requirements.  
+
+---
+
+## Limitations
+
+- **Numeric labels only:** Original training dataset did not include human readable genre names.  
+- **Collapsed predictions:** Model tends to overpredict the majority class (`0`).  
+- **Generalization:** Accuracy on unseen books is uncertain due to limited feature set.  
+
+---
+
+## Future Improvements
+
+- Map numeric codes to the actual genre categories from the dataset.  
+- Retrain model with balanced classes.  
+- Provide confidence scores along with predictions.  
+- Explore richer book features (author, publisher, language).  
 
 ---
 
-## How This Fits the Assignment
+## AI Disclosure
+
+Parts of this project were supported with the help of AI tools (GPT-5), mainly for:  
+- Debugging deployment issues on Hugging Face Spaces  
+- Improving the stability of the Gradio interface  
+- Polishing documentation  
 
-- ✅ Uses a **classmate’s tabular model** (not my own).  
-- ✅ Researched and linked the **dataset/model docs**.  
-- ✅ Built a Gradio app with **widgets + examples**.  
-- ✅ Exposed inputs with validation and presented predictions clearly.  
-- ✅ Deployed publicly on Hugging Face Spaces with proper documentation.  
+The dataset, model training, and integration choices remain based on classmate provided artifacts and my own implementation work.
 
 ---