README.md

---
language: en
license: mit
tags:
  - image-classification
  - tensorflow
  - balls-dataset
  - image-processing
inference: true
datasets:
  - AIOmarRehan/Sports-Balls
---

# Sports_Balls_Classification.h5

## Model Details

This is a trained InceptionV3 transfer learning model for classifying 15 different types of sports balls.

## Specifications

- Architecture: InceptionV3 with custom classification head
- Input Size: 225 x 225 pixels (RGB)
- Output Classes: 15 sports ball types
- Framework: TensorFlow/Keras
- Format: H5 (HDF5)

## Supported Sports Ball Types

1. American Football
2. Baseball
3. Basketball
4. Billiard Ball
5. Bowling Ball
6. Cricket Ball
7. Football
8. Golf Ball
9. Hockey Ball
10. Hockey Puck
11. Rugby Ball
12. Shuttlecock
13. Table Tennis Ball
14. Tennis Ball
15. Volleyball

## Loading and Using

### Python Example

```python
import tensorflow as tf
from PIL import Image
import numpy as np

model = tf.keras.models.load_model("Sports_Balls_Classification.h5")

img = Image.open("sports_ball.jpg").convert("RGB")
img = img.resize((225, 225))
img_array = np.array(img).astype("float32") / 255.0
img_array = np.expand_dims(img_array, axis=0)

predictions = model.predict(img_array)
predicted_class = np.argmax(predictions[0])
confidence = np.max(predictions[0])
```

## Training Approach

- Stage 1: Feature extraction (5 epochs) - Base frozen
- Stage 2: Fine-tuning (10 epochs) - Last 30 layers unfrozen
- Data balancing: 808 images per class
- Callbacks: Early stopping, learning rate reduction, checkpointing

## Performance

Trained on balanced, preprocessed sports ball images with augmentation.
Achieves high accuracy across all 15 sports ball classes.

## Requirements

- TensorFlow >= 2.0
- Pillow
- NumPy

## License

MIT