SAR Image Colorization — U-Net FastAPI

Dockerfile

@@ -0,0 +1,16 @@
+FROM python:3.10-slim
+
+RUN useradd -m -u 1000 user
+USER user
+ENV PATH="/home/user/.local/bin:$PATH"
+
+WORKDIR /app
+
+COPY --chown=user requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY --chown=user . .
+
+RUN mkdir -p static
+
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -0,0 +1,14 @@
+# SAR Image Colorization
+
+Convert grayscale Sentinel-1 SAR radar images to optical color using U-Net deep learning.
+
+## Model
+- Architecture: U-Net (encoder-decoder with skip connections)
+- Input: 256x256 grayscale SAR image
+- Output: 256x256 RGB colorized image
+- Dataset: 16,000 paired Sentinel-1/Sentinel-2 images
+
+## Tech Stack
+- TensorFlow 2.16 + U-Net
+- FastAPI + Uvicorn
+- Deployed on Hugging Face Spaces

main.py

@@ -0,0 +1,117 @@
+from fastapi import FastAPI, File, UploadFile, HTTPException
+from fastapi.staticfiles import StaticFiles
+from fastapi.templating import Jinja2Templates
+from fastapi.requests import Request
+from fastapi.responses import JSONResponse
+import numpy as np
+from PIL import Image
+import io
+import os
+import base64
+import tensorflow as tf
+from tensorflow.keras import layers, Model
+
+app = FastAPI(title="SAR Image Colorization")
+
+app.mount("/static", StaticFiles(directory="static"), name="static")
+templates = Jinja2Templates(directory="templates")
+
+IMG_SIZE   = (256, 256)
+MODEL_PATH = "sar_model.weights.h5"
+model      = None
+
+def build_unet(input_shape=(256, 256, 1)):
+    inputs = layers.Input(input_shape)
+    def conv_block(x, filters):
+        x = layers.Conv2D(filters, 3, padding="same", activation="relu")(x)
+        x = layers.BatchNormalization()(x)
+        x = layers.Conv2D(filters, 3, padding="same", activation="relu")(x)
+        x = layers.BatchNormalization()(x)
+        return x
+    def encoder_block(x, filters):
+        skip = conv_block(x, filters)
+        pool = layers.MaxPooling2D(2)(skip)
+        return skip, pool
+    def decoder_block(x, skip, filters):
+        x = layers.Conv2DTranspose(filters, 2, strides=2, padding="same")(x)
+        x = layers.Concatenate()([x, skip])
+        x = conv_block(x, filters)
+        return x
+    s1, p1 = encoder_block(inputs, 32)
+    s2, p2 = encoder_block(p1,     64)
+    s3, p3 = encoder_block(p2,     128)
+    s4, p4 = encoder_block(p3,     256)
+    b      = conv_block(p4, 512)
+    d1     = decoder_block(b,  s4, 256)
+    d2     = decoder_block(d1, s3, 128)
+    d3     = decoder_block(d2, s2, 64)
+    d4     = decoder_block(d3, s1, 32)
+    outputs = layers.Conv2D(3, 1, activation="tanh")(d4)
+    return Model(inputs, outputs)
+
+@app.on_event("startup")
+async def load_model():
+    global model
+    if os.path.exists(MODEL_PATH):
+        model = build_unet()
+        model(tf.zeros((1, 256, 256, 1)))  # build model
+        model.load_weights(MODEL_PATH)
+        print("✅ SAR model loaded successfully")
+    else:
+        print("⚠️  Model not found — demo mode")
+
+def preprocess(image_bytes):
+    img = Image.open(io.BytesIO(image_bytes)).convert("L")
+    # Use LANCZOS for high quality resize
+    img = img.resize(IMG_SIZE, Image.LANCZOS)
+    arr = np.array(img, dtype=np.float32) / 127.5 - 1.0
+    return np.expand_dims(arr[..., np.newaxis], 0)
+
+def to_base64(arr_uint8):
+    img = Image.fromarray(arr_uint8)
+    buf = io.BytesIO()
+    img.save(buf, format="PNG")
+    return base64.b64encode(buf.getvalue()).decode()
+
+@app.get("/")
+async def home(request: Request):
+    return templates.TemplateResponse("index.html", {"request": request})
+
+@app.post("/colorize")
+async def colorize(file: UploadFile = File(...)):
+    if not file.content_type.startswith("image/"):
+        raise HTTPException(status_code=400, detail="File must be an image.")
+    contents = await file.read()
+    if len(contents) > 10 * 1024 * 1024:
+        raise HTTPException(status_code=400, detail="Image too large. Max 10MB.")
+    try:
+        inp = preprocess(contents)
+    except Exception:
+        raise HTTPException(status_code=400, detail="Could not process image.")
+
+    if model is None:
+        # Demo mode
+        import random
+        dummy = np.random.randint(50, 200, (256, 256, 3), dtype=np.uint8)
+        dummy[:,:,0] = np.clip(dummy[:,:,0], 30, 100)
+        dummy[:,:,1] = np.clip(dummy[:,:,1], 80, 180)
+        dummy[:,:,2] = np.clip(dummy[:,:,2], 30, 100)
+        pred_b64 = to_base64(dummy)
+    else:
+        pred = model.predict(inp, verbose=0)[0]
+        pred_uint8 = ((pred + 1) * 127.5).clip(0, 255).astype(np.uint8)
+        pred_b64 = to_base64(pred_uint8)
+
+    # Also return input as base64 for display
+    inp_disp = ((inp[0,:,:,0] + 1) * 127.5).clip(0, 255).astype(np.uint8)
+    inp_b64  = to_base64(inp_disp)
+
+    return JSONResponse({
+        "success":    True,
+        "input_b64":  inp_b64,
+        "output_b64": pred_b64
+    })
+
+@app.get("/health")
+async def health():
+    return {"status": "ok", "model_loaded": model is not None}

requirements.txt

@@ -0,0 +1,7 @@
+fastapi==0.104.1
+uvicorn[standard]==0.24.0
+python-multipart==0.0.6
+jinja2==3.1.2
+Pillow==10.1.0
+numpy==1.26.4
+tensorflow==2.16.1

templates/index.html

@@ -0,0 +1,253 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>SAR Image Colorization</title>
+<style>
+  * { margin: 0; padding: 0; box-sizing: border-box; }
+  body { background: #0d1117; color: #e6edf3; font-family: Arial, sans-serif; min-height: 100vh; }
+
+  header {
+    background: linear-gradient(135deg, #1a2a4a, #0d1117);
+    padding: 24px 40px;
+    border-bottom: 1px solid #30363d;
+  }
+  header h1 { font-size: 28px; color: #58a6ff; }
+  header p  { color: #8b949e; font-size: 14px; margin-top: 4px; }
+  .tag { display: inline-block; background: #1f6feb33; color: #58a6ff; border: 1px solid #1f6feb; border-radius: 20px; padding: 2px 12px; font-size: 12px; margin-top: 8px; margin-right: 6px; }
+
+  .container { max-width: 1100px; margin: 40px auto; padding: 0 24px; }
+
+  .upload-section {
+    background: #161b22;
+    border: 2px dashed #30363d;
+    border-radius: 12px;
+    padding: 40px;
+    text-align: center;
+    cursor: pointer;
+    transition: border-color 0.3s;
+    margin-bottom: 32px;
+  }
+  .upload-section:hover { border-color: #58a6ff; }
+  .upload-section.dragover { border-color: #58a6ff; background: #1f2937; }
+  .upload-icon { font-size: 48px; margin-bottom: 12px; }
+  .upload-section h2 { font-size: 20px; margin-bottom: 8px; }
+  .upload-section p  { color: #8b949e; font-size: 14px; }
+
+  input[type="file"] { display: none; }
+
+  .btn {
+    background: linear-gradient(135deg, #1f6feb, #58a6ff);
+    color: white;
+    border: none;
+    padding: 12px 32px;
+    border-radius: 8px;
+    font-size: 16px;
+    cursor: pointer;
+    margin-top: 16px;
+    transition: opacity 0.2s;
+  }
+  .btn:hover { opacity: 0.85; }
+  .btn:disabled { opacity: 0.4; cursor: not-allowed; }
+
+  .result-section { display: none; }
+  .result-section h2 { font-size: 22px; margin-bottom: 20px; color: #58a6ff; }
+
+  .images-grid {
+    display: grid;
+    grid-template-columns: 1fr 60px 1fr;
+    gap: 0;
+    align-items: center;
+    margin-bottom: 32px;
+  }
+  .image-card {
+    background: #161b22;
+    border: 1px solid #30363d;
+    border-radius: 12px;
+    overflow: hidden;
+  }
+  .image-card .label {
+    padding: 12px 16px;
+    font-size: 13px;
+    font-weight: bold;
+    border-bottom: 1px solid #30363d;
+  }
+  .image-card.input .label  { color: #8b949e; }
+  .image-card.output .label { color: #3fb950; }
+  .image-card img {
+    width: 100%;
+    display: block;
+    image-rendering: pixelated;
+  }
+
+  .arrow {
+    text-align: center;
+    font-size: 36px;
+    color: #58a6ff;
+  }
+
+  .loading {
+    display: none;
+    text-align: center;
+    padding: 40px;
+  }
+  .spinner {
+    width: 48px; height: 48px;
+    border: 4px solid #30363d;
+    border-top-color: #58a6ff;
+    border-radius: 50%;
+    animation: spin 0.8s linear infinite;
+    margin: 0 auto 16px;
+  }
+  @keyframes spin { to { transform: rotate(360deg); } }
+
+  .how-it-works {
+    background: #161b22;
+    border: 1px solid #30363d;
+    border-radius: 12px;
+    padding: 24px;
+    margin-top: 32px;
+  }
+  .how-it-works h3 { color: #58a6ff; margin-bottom: 16px; }
+  .steps { display: grid; grid-template-columns: repeat(4, 1fr); gap: 16px; }
+  .step { text-align: center; }
+  .step-num { width: 36px; height: 36px; border-radius: 50%; background: #1f6feb; color: white; font-weight: bold; display: flex; align-items: center; justify-content: center; margin: 0 auto 8px; }
+  .step p { font-size: 13px; color: #8b949e; }
+
+  footer {
+    text-align: center;
+    padding: 24px;
+    color: #8b949e;
+    font-size: 13px;
+    border-top: 1px solid #30363d;
+    margin-top: 48px;
+  }
+</style>
+</head>
+<body>
+
+<header>
+  <h1>🛰️ SAR Image Colorization</h1>
+  <p>Convert grayscale Sentinel-1 SAR radar images to optical color using U-Net deep learning</p>
+  <span class="tag">U-Net</span>
+  <span class="tag">Sentinel-1 → Sentinel-2</span>
+  <span class="tag">Image-to-Image Translation</span>
+</header>
+
+<div class="container">
+
+  <div class="upload-section" id="dropZone" onclick="document.getElementById('fileInput').click()">
+    <div class="upload-icon">🛰️</div>
+    <h2>Upload SAR Image</h2>
+    <p>Drag & drop a Sentinel-1 grayscale image or click to browse</p>
+    <p style="margin-top:8px; font-size:12px; color:#6e7681;">Supports PNG, JPG • Max 10MB</p>
+    <button class="btn" onclick="event.stopPropagation(); document.getElementById('fileInput').click()">Choose Image</button>
+    <input type="file" id="fileInput" accept="image/*">
+  </div>
+
+  <div class="loading" id="loading">
+    <div class="spinner"></div>
+    <p style="color:#8b949e;">Colorizing your SAR image...</p>
+  </div>
+
+  <div class="result-section" id="resultSection">
+    <h2>✅ Colorization Result</h2>
+    <div class="images-grid">
+      <div class="image-card input">
+        <div class="label">📡 SAR Input (Grayscale)</div>
+        <img id="inputImg" src="" alt="Input">
+      </div>
+      <div class="arrow">→</div>
+      <div class="image-card output">
+        <div class="label">🌍 Colorized Output (Predicted)</div>
+        <img id="outputImg" src="" alt="Output">
+      </div>
+    </div>
+    <button class="btn" onclick="resetApp()">Try Another Image</button>
+  </div>
+
+  <div class="how-it-works">
+    <h3>How It Works</h3>
+    <div class="steps">
+      <div class="step">
+        <div class="step-num">1</div>
+        <p>Upload a grayscale SAR radar image from Sentinel-1</p>
+      </div>
+      <div class="step">
+        <div class="step-num">2</div>
+        <p>U-Net encoder extracts spatial features at multiple scales</p>
+      </div>
+      <div class="step">
+        <div class="step-num">3</div>
+        <p>Decoder reconstructs image with predicted RGB colors</p>
+      </div>
+      <div class="step">
+        <div class="step-num">4</div>
+        <p>Output resembles what a Sentinel-2 optical satellite would see</p>
+      </div>
+    </div>
+  </div>
+
+</div>
+
+<footer>
+  Built by Sanjay S &nbsp;|&nbsp; U-Net · TensorFlow · FastAPI &nbsp;|&nbsp;
+  <a href="https://github.com/sanjay5656/sar-colorization" style="color:#58a6ff;">GitHub</a>
+</footer>
+
+<script>
+const dropZone   = document.getElementById("dropZone");
+const fileInput  = document.getElementById("fileInput");
+const loading    = document.getElementById("loading");
+const resultSec  = document.getElementById("resultSection");
+const inputImg   = document.getElementById("inputImg");
+const outputImg  = document.getElementById("outputImg");
+
+fileInput.addEventListener("change", e => {
+  if (e.target.files[0]) handleFile(e.target.files[0]);
+});
+
+dropZone.addEventListener("dragover", e => {
+  e.preventDefault();
+  dropZone.classList.add("dragover");
+});
+dropZone.addEventListener("dragleave", () => dropZone.classList.remove("dragover"));
+dropZone.addEventListener("drop", e => {
+  e.preventDefault();
+  dropZone.classList.remove("dragover");
+  if (e.dataTransfer.files[0]) handleFile(e.dataTransfer.files[0]);
+});
+
+async function handleFile(file) {
+  dropZone.style.display  = "none";
+  loading.style.display   = "block";
+  resultSec.style.display = "none";
+
+  const formData = new FormData();
+  formData.append("file", file);
+
+  try {
+    const res  = await fetch("/colorize", { method: "POST", body: formData });
+    const data = await res.json();
+    if (data.success) {
+      inputImg.src  = "data:image/png;base64," + data.input_b64;
+      outputImg.src = "data:image/png;base64," + data.output_b64;
+      loading.style.display   = "none";
+      resultSec.style.display = "block";
+    }
+  } catch (err) {
+    alert("Error: " + err.message);
+    resetApp();
+  }
+}
+
+function resetApp() {
+  dropZone.style.display  = "block";
+  loading.style.display   = "none";
+  resultSec.style.display = "none";
+  fileInput.value = "";
+}
+</script>
+</body>
+</html>