Spaces:

lablab-ai-amd-developer-hackathon
/

urban-expansion-detector

Running

File size: 9,333 Bytes

import gradio as gr
import requests
import base64
from PIL import Image
from io import BytesIO
import tempfile
import os
 
API_BASE = "http://134.199.195.79:8001"
 
 
def b64_to_pil(b64_str):
    return Image.open(BytesIO(base64.b64decode(b64_str)))
 
 
def analyze_single(image):
    if image is None:
        return None, "⚠ Upload a satellite tile first."
    buf = BytesIO()
    image.save(buf, format="PNG")
    buf.seek(0)
    try:
        r = requests.post(
            f"{API_BASE}/analyze",
            files={"file": ("tile.png", buf, "image/png")},
            timeout=300,
        )
        r.raise_for_status()
        data = r.json()
        annotated = b64_to_pil(data["annotated_image"])
        built = data.get("built_fraction")
        built_str = f"{built:.1%}" if built else "—"
        boxes = data.get("boxes_found", 0)
        header = f"**Built area:** {built_str}  |  **Urban clusters detected:** {boxes}\n\n---\n\n"
        return annotated, header + data["analysis"]
    except Exception as e:
        return None, f"❌ Error: {e}"
 
 
def analyze_corridor(images):
    if not images:
        return [], "⚠ Upload at least 2 tiles.", None
    files = []
    for i, img in enumerate(images):
        buf = BytesIO()
        if isinstance(img, tuple):
            img = img[0]
        if isinstance(img, str):
            img = Image.open(img)
        img.save(buf, format="PNG")
        files.append(("files", (f"tile_{i+1}.png", buf.getvalue(), "image/png")))
    try:
        r = requests.post(
            f"{API_BASE}/corridor-report",
            files=files,
            timeout=900,
        )
        r.raise_for_status()
        data = r.json()
        annotated_pils = [b64_to_pil(b) for b in data["annotated_images"]]
        md = f"## 🛰 Corridor Summary\n\n{data['corridor_summary']}\n\n---\n\n"
        for t in data["tiles"]:
            frac = t["built_fraction"]
            frac_str = f"{frac:.1%}" if frac else "—"
            md += f"### Tile {t['tile']}  ·  Built: {frac_str}\n{t['analysis']}\n\n"
        gallery_output = [(img, f"Tile {i+1}") for i, img in enumerate(annotated_pils)]
        return gallery_output, md, data.get("pdf_path")
    except Exception as e:
        return [], f"❌ Error: {e}", None
 
 
def download_pdf(pdf_path):
    if not pdf_path:
        return None
    try:
        r = requests.get(
            f"{API_BASE}/download-report",
            params={"path": pdf_path},
            timeout=60,
        )
        r.raise_for_status()
        tmp = tempfile.NamedTemporaryFile(suffix=".pdf", delete=False)
        tmp.write(r.content)
        tmp.flush()
        return tmp.name
    except Exception as e:
        return None
 
 
CSS = """
@import url('https://fonts.googleapis.com/css2?family=Syne:wght@400;700;800&family=JetBrains+Mono:wght@400;500&display=swap');
:root {
    --bg: #060a06;
    --surface: #0d150d;
    --border: #1a2e1a;
    --accent: #00e676;
    --accent2: #69ff47;
    --text: #d4ecd4;
    --muted: #5a7a5a;
    --danger: #ff5252;
}
body, .gradio-container {
    background: var(--bg) !important;
    font-family: 'Syne', sans-serif !important;
    color: var(--text) !important;
}
.app-header {
    border-bottom: 1px solid var(--border);
    padding: 2rem 0 1.5rem;
    margin-bottom: 1.5rem;
}
.app-title {
    font-size: 2.2rem;
    font-weight: 800;
    color: var(--accent);
    letter-spacing: -0.03em;
    line-height: 1;
    margin: 0;
}
.app-sub {
    font-family: 'JetBrains Mono', monospace;
    font-size: 0.72rem;
    color: var(--muted);
    margin-top: 0.4rem;
    letter-spacing: 0.08em;
}
.tab-nav {
    background: var(--surface) !important;
    border: 1px solid var(--border) !important;
    border-radius: 6px !important;
    gap: 2px !important;
    padding: 3px !important;
}
.tab-nav button {
    font-family: 'JetBrains Mono', monospace !important;
    font-size: 0.78rem !important;
    color: var(--muted) !important;
    background: transparent !important;
    border-radius: 4px !important;
    padding: 6px 18px !important;
    letter-spacing: 0.05em;
    transition: all 0.15s;
}
.tab-nav button.selected {
    background: var(--accent) !important;
    color: #000 !important;
    font-weight: 700 !important;
}
button.primary {
    background: var(--accent) !important;
    color: #000 !important;
    font-family: 'JetBrains Mono', monospace !important;
    font-weight: 700 !important;
    font-size: 0.82rem !important;
    letter-spacing: 0.06em !important;
    border: none !important;
    border-radius: 4px !important;
    padding: 10px 24px !important;
    transition: opacity 0.15s !important;
}
button.primary:hover { opacity: 0.85 !important; }
button.secondary {
    background: transparent !important;
    color: var(--accent) !important;
    font-family: 'JetBrains Mono', monospace !important;
    font-size: 0.78rem !important;
    border: 1px solid var(--accent) !important;
    border-radius: 4px !important;
    padding: 8px 20px !important;
    transition: all 0.15s !important;
}
button.secondary:hover {
    background: var(--accent) !important;
    color: #000 !important;
}
.panel {
    background: var(--surface) !important;
    border: 1px solid var(--border) !important;
    border-radius: 8px !important;
}
label span {
    font-family: 'JetBrains Mono', monospace !important;
    font-size: 0.72rem !important;
    color: var(--muted) !important;
    letter-spacing: 0.08em !important;
    text-transform: uppercase !important;
}
textarea, .prose {
    font-family: 'JetBrains Mono', monospace !important;
    font-size: 0.82rem !important;
    background: var(--bg) !important;
    color: var(--text) !important;
    border: 1px solid var(--border) !important;
    border-radius: 4px !important;
}
.footer-strip {
    font-family: 'JetBrains Mono', monospace;
    font-size: 0.68rem;
    color: var(--muted);
    border-top: 1px solid var(--border);
    padding-top: 1rem;
    margin-top: 2rem;
    letter-spacing: 0.04em;
}
"""
 
with gr.Blocks(title="Urban Expansion Detector") as demo:
 
    pdf_path_state = gr.State(None)
 
    gr.HTML("""
    <div class="app-header">
        <p class="app-title">🛰 URBAN EXPANSION DETECTOR</p>
        <p class="app-sub">
            QWEN2.5-VL 72B · LORA FINE-TUNE · AMD MI300X · 8,000 SENTINEL-2 TILES
        </p>
    </div>
    """)
 
    with gr.Tabs():
 
        with gr.Tab("SINGLE TILE"):
            with gr.Row(equal_height=True):
                with gr.Column(scale=1, elem_classes="panel"):
                    single_in = gr.Image(
                        label="INPUT — SATELLITE TILE",
                        type="pil",
                        height=340,
                    )
                    single_btn = gr.Button("ANALYZE →", variant="primary")
                with gr.Column(scale=1, elem_classes="panel"):
                    single_out_img = gr.Image(
                        label="OUTPUT — ANNOTATED",
                        height=340,
                        interactive=False,
                    )
            single_out_text = gr.Textbox(
                label="MODEL ANALYSIS",
                lines=7,
                elem_classes="panel",
            )
            single_btn.click(
                analyze_single,
                inputs=[single_in],
                outputs=[single_out_img, single_out_text],
            )
 
        with gr.Tab("CORRIDOR ANALYSIS"):
            gr.HTML("""
            <p style="font-family:'JetBrains Mono',monospace;font-size:0.78rem;
                      color:#5a7a5a;margin-bottom:1rem;">
                Upload 2–6 tiles along a corridor (e.g. Delhi–Meerut RRTS).
                Each tile is analyzed independently, then a corridor-level summary is generated.
            </p>
            """)
            corridor_in = gr.Gallery(
                label="INPUT — CORRIDOR TILES",
                type="pil",
                columns=3,
                height=260,
                elem_classes="panel",
            )
            corridor_btn = gr.Button("ANALYZE CORRIDOR →", variant="primary")
            corridor_out_gallery = gr.Gallery(
                label="OUTPUT — ANNOTATED TILES",
                columns=3,
                height=300,
                elem_classes="panel",
            )
            corridor_out_text = gr.Markdown(
                label="CORRIDOR SUMMARY",
                elem_classes="panel",
            )
            with gr.Row():
                pdf_btn = gr.Button("📄 EXPORT PDF REPORT", variant="secondary")
                pdf_out = gr.File(label="DOWNLOAD", elem_classes="panel")
 
            corridor_btn.click(
                analyze_corridor,
                inputs=[corridor_in],
                outputs=[corridor_out_gallery, corridor_out_text, pdf_path_state],
            )
            pdf_btn.click(
                download_pdf,
                inputs=[pdf_path_state],
                outputs=[pdf_out],
            )
 
    gr.HTML("""
    <div class="footer-strip">
        MODEL · MohitML10/urban-expansion-detector-72b-v3 &nbsp;·&nbsp;
        HARDWARE · AMD MI300X 192GB HBM3 &nbsp;·&nbsp;
        DATA · 8,000 SENTINEL-2 TILES (437 INDIA) &nbsp;·&nbsp;
        USE CASE · URBAN PLANNING · RRTS FEASIBILITY
    </div>
    """)
 
demo.launch(css=CSS)