example.py

"""
Shape Polygons Dataset - Usage Examples

This script demonstrates various ways to load and use the Shape Polygons Dataset.
"""

import os
import pandas as pd
import matplotlib.pyplot as plt
from PIL import Image

# =============================================================================
# Basic Usage: Loading Metadata and Images
# =============================================================================

def load_dataset(data_dir=".", split="train"):
    """Load metadata and return as pandas DataFrame."""
    metadata_path = os.path.join(data_dir, split, "metadata.csv")
    return pd.read_csv(metadata_path)


def load_image(data_dir, split, filename):
    """Load a single image from the dataset."""
    img_path = os.path.join(data_dir, split, "images", filename)
    return Image.open(img_path)


# =============================================================================
# Example 1: Explore Dataset Statistics
# =============================================================================

def explore_statistics(data_dir="."):
    """Print dataset statistics."""
    print("=" * 50)
    print("Shape Polygons Dataset Statistics")
    print("=" * 50)
    
    for split in ["train", "test"]:
        df = load_dataset(data_dir, split)
        print(f"\n{split.upper()} Split:")
        print(f"  Total images: {len(df)}")
        print(f"\n  Vertices distribution:")
        for v in range(3, 9):
            count = len(df[df["vertices"] == v])
            print(f"    {v} vertices: {count} ({count/len(df)*100:.1f}%)")
        
        print(f"\n  Size statistics:")
        print(f"    Min: {df['size'].min():.4f}")
        print(f"    Max: {df['size'].max():.4f}")
        print(f"    Mean: {df['size'].mean():.4f}")


# =============================================================================
# Example 2: Visualize Sample Images
# =============================================================================

def visualize_samples(data_dir=".", n_samples=12, split="train"):
    """Visualize random samples from the dataset."""
    df = load_dataset(data_dir, split)
    samples = df.sample(n=min(n_samples, len(df)))
    
    n_cols = 4
    n_rows = (len(samples) + n_cols - 1) // n_cols
    
    fig, axes = plt.subplots(n_rows, n_cols, figsize=(12, 3 * n_rows))
    axes = axes.flatten() if n_samples > 1 else [axes]
    
    for idx, (_, row) in enumerate(samples.iterrows()):
        img = load_image(data_dir, split, row["filename"])
        axes[idx].imshow(img)
        axes[idx].set_title(f"{row['vertices']} vertices\nsize={row['size']:.2f}")
        axes[idx].axis("off")
    
    # Hide empty subplots
    for idx in range(len(samples), len(axes)):
        axes[idx].axis("off")
    
    plt.tight_layout()
    plt.savefig("samples_visualization.png", dpi=150, bbox_inches="tight")
    print(f"Saved visualization to 'samples_visualization.png'")
    plt.show()


# =============================================================================
# Example 3: Visualize by Shape Type
# =============================================================================

def visualize_by_shape_type(data_dir=".", split="train"):
    """Show one example of each shape type."""
    df = load_dataset(data_dir, split)
    shape_names = {
        3: "Triangle",
        4: "Quadrilateral",
        5: "Pentagon",
        6: "Hexagon",
        7: "Heptagon",
        8: "Octagon"
    }
    
    fig, axes = plt.subplots(2, 3, figsize=(12, 8))
    axes = axes.flatten()
    
    for idx, vertices in enumerate(range(3, 9)):
        sample = df[df["vertices"] == vertices].iloc[0]
        img = load_image(data_dir, split, sample["filename"])
        axes[idx].imshow(img)
        axes[idx].set_title(f"{shape_names[vertices]}\n({vertices} vertices)")
        axes[idx].axis("off")
    
    plt.suptitle("Shape Types in Dataset", fontsize=14, fontweight="bold")
    plt.tight_layout()
    plt.savefig("shape_types.png", dpi=150, bbox_inches="tight")
    print(f"Saved visualization to 'shape_types.png'")
    plt.show()


# =============================================================================
# Example 4: PyTorch Dataset Class
# =============================================================================

# Optional imports for PyTorch functionality
try:
    import torch
    from torch.utils.data import Dataset, DataLoader
    from torchvision import transforms
    PYTORCH_AVAILABLE = True
except ImportError:
    PYTORCH_AVAILABLE = False


class ShapePolygonsDataset:
    """PyTorch Dataset for Shape Polygons.
    
    Requires: torch, torchvision
    """
    
    def __init__(self, root_dir, split="train", transform=None, task="classification"):
        """
        Args:
            root_dir: Root directory of the dataset
            split: "train" or "test"
            transform: Optional torchvision transforms
            task: "classification" for vertex count, "regression" for size prediction, "multi" for all properties
        """
        if not PYTORCH_AVAILABLE:
            raise ImportError("PyTorch is required. Install with: pip install torch torchvision")
        
        self.root_dir = root_dir
        self.split = split
        self.transform = transform
        self.task = task
        self.metadata = pd.read_csv(os.path.join(root_dir, split, "metadata.csv"))
        
        if self.transform is None:
            self.transform = transforms.Compose([
                transforms.ToTensor(),
                transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
            ])
    
    def __len__(self):
        return len(self.metadata)
    
    def __getitem__(self, idx):
        row = self.metadata.iloc[idx]
        img_path = os.path.join(self.root_dir, self.split, "images", row["filename"])
        image = Image.open(img_path).convert("RGB")
        
        if self.transform:
            image = self.transform(image)
        
        if self.task == "classification":
            # Class label: 0-5 for 3-8 vertices
            label = torch.tensor(row["vertices"] - 3, dtype=torch.long)
        elif self.task == "regression":
            # Predict size
            label = torch.tensor(row["size"], dtype=torch.float32)
        elif self.task == "multi":
            # Multi-task: return all properties
            label = {
                "vertices": torch.tensor(row["vertices"] - 3, dtype=torch.long),
                "size": torch.tensor(row["size"], dtype=torch.float32),
                "angle": torch.tensor(row["angle"], dtype=torch.float32),
                "center": torch.tensor([row["center_x"], row["center_y"]], dtype=torch.float32),
                "color": torch.tensor([row["color_r"], row["color_g"], row["color_b"]], dtype=torch.float32)
            }
        else:
            raise ValueError(f"Unknown task: {self.task}")
        
        return image, label


def demo_pytorch_dataloader(data_dir="."):
    """Demonstrate PyTorch DataLoader usage."""
    if not PYTORCH_AVAILABLE:
        print("PyTorch is not installed. Install with: pip install torch torchvision")
        return
    
    print("Creating PyTorch Dataset and DataLoader...")
    dataset = ShapePolygonsDataset(data_dir, split="train", task="classification")
    dataloader = DataLoader(dataset, batch_size=32, shuffle=True, num_workers=0)
    
    # Get one batch
    images, labels = next(iter(dataloader))
    print(f"Batch shape: {images.shape}")
    print(f"Labels shape: {labels.shape}")
    print(f"Label values (vertices - 3): {labels[:10].tolist()}")
    print(f"Actual vertex counts: {[l + 3 for l in labels[:10].tolist()]}")


# =============================================================================
# Example 5: Color Analysis
# =============================================================================

def analyze_colors(data_dir=".", split="train"):
    """Analyze color distribution in the dataset."""
    df = load_dataset(data_dir, split)
    
    fig, axes = plt.subplots(1, 3, figsize=(15, 4))
    colors = ["red", "green", "blue"]
    columns = ["color_r", "color_g", "color_b"]
    
    for idx, (color, col) in enumerate(zip(colors, columns)):
        axes[idx].hist(df[col], bins=50, color=color, alpha=0.7, edgecolor="black")
        axes[idx].set_xlabel(f"{color.capitalize()} Value")
        axes[idx].set_ylabel("Frequency")
        axes[idx].set_title(f"{color.capitalize()} Channel Distribution")
    
    plt.suptitle(f"Color Distribution in {split.capitalize()} Set", fontsize=14, fontweight="bold")
    plt.tight_layout()
    plt.savefig("color_distribution.png", dpi=150, bbox_inches="tight")
    print(f"Saved visualization to 'color_distribution.png'")
    plt.show()


# =============================================================================
# Main
# =============================================================================

if __name__ == "__main__":
    import argparse
    
    parser = argparse.ArgumentParser(description="Shape Polygons Dataset Examples")
    parser.add_argument("--data-dir", type=str, default=".", help="Path to dataset root")
    parser.add_argument(
        "--example",
        type=str,
        choices=["stats", "samples", "shapes", "pytorch", "colors", "all"],
        default="all",
        help="Which example to run"
    )
    args = parser.parse_args()
    
    examples = {
        "stats": ("Dataset Statistics", lambda: explore_statistics(args.data_dir)),
        "samples": ("Sample Visualization", lambda: visualize_samples(args.data_dir)),
        "shapes": ("Shape Types", lambda: visualize_by_shape_type(args.data_dir)),
        "pytorch": ("PyTorch DataLoader Demo", lambda: demo_pytorch_dataloader(args.data_dir)),
        "colors": ("Color Analysis", lambda: analyze_colors(args.data_dir)),
    }
    
    if args.example == "all":
        for name, (desc, func) in examples.items():
            print(f"\n{'=' * 50}")
            print(f"Example: {desc}")
            print("=" * 50)
            func()
    else:
        name = args.example
        desc, func = examples[name]
        print(f"Running Example: {desc}")
        func()