Initial commit

app.py

@@ -0,0 +1,72 @@
+import gradio as gr
+import numpy as np
+from sklearn.cluster import KMeans
+from PIL import Image
+
+def compress_kmeans(image: np.ndarray, k: int) -> np.ndarray:
+    """Reduce image palette to k colours using K-Means vector quantisation."""
+    pixels = image.reshape(-1, 3)
+    model = KMeans(n_clusters=k, n_init="auto", random_state=42)
+    model.fit(pixels)
+    centres = model.cluster_centers_.astype(np.uint8)
+    labels = model.predict(pixels)
+    return centres[labels].reshape(image.shape)
+
+def compute_psnr_mse(original: np.ndarray, compressed: np.ndarray):
+    """Return (MSE, PSNR) between two images."""
+    mse = np.mean((original.astype(np.float64) - compressed.astype(np.float64)) ** 2)
+    psnr = float("inf") if mse == 0 else 20.0 * np.log10(255.0 / np.sqrt(mse))
+    return mse, psnr
+
+def compute_metrics(image: np.ndarray, k: int):
+    """Estimate compression ratio and file sizes (theoretical)."""
+    h, w, c = image.shape
+    original_bits = h * w * c * 8
+    compressed_bits = (k * 24) + (h * w * int(np.ceil(np.log2(k))))
+    ratio = original_bits / compressed_bits
+    return {
+        "ratio": ratio,
+        "original_kb": original_bits / (8 * 1024),
+        "compressed_kb": compressed_bits / (8 * 1024),
+        "saved_percent": (1.0 - (1.0 / ratio)) * 100.0,
+    }
+
+def compress_and_report(image, k):
+    """Gradio entry point: compress, compute metrics, return image + report."""
+    if isinstance(image, np.ndarray):
+        orig = image[:, :, :3] if image.shape[-1] == 4 else image
+    else:
+        orig = np.array(image.convert("RGB"))
+
+    compressed = compress_kmeans(orig, k)
+    mse, psnr = compute_psnr_mse(orig, compressed)
+    metrics = compute_metrics(orig, k)
+
+    report = (
+        f"COMPRESSION REPORT\n"
+        f"{'-' * 40}\n"
+        f"PSNR               : {psnr:.2f} dB\n"
+        f"MSE                : {mse:.2f}\n"
+        f"Compression Ratio  : {metrics['ratio']:.2f}x\n"
+        f"Original Size (KB) : {metrics['original_kb']:.2f}\n"
+        f"Compressed Size (KB): {metrics['compressed_kb']:.2f}\n"
+        f"Space Saved (%)    : {metrics['saved_percent']:.1f}"
+    )
+    return compressed, report
+
+iface = gr.Interface(
+    fn=compress_and_report,
+    inputs=[
+        gr.Image(type="pil", label="Upload Image"),
+        gr.Slider(minimum=2, maximum=64, step=2, value=16, label="Number of Colors (k)"),
+    ],
+    outputs=[
+        gr.Image(type="numpy", label="Compressed Image"),
+        gr.Textbox(label="Metrics Report", lines=10),
+    ],
+    title="Image Compression using K-Means",
+    description="Upload an image and adjust the colour palette size to see lossy compression in action.",
+)
+
+if __name__ == "__main__":
+    iface.launch()

requirements.txt

@@ -0,0 +1,4 @@
+gradio
+scikit-learn
+numpy
+Pillow