| import json |
| import os |
| import random |
| from collections import Counter, defaultdict |
| from datetime import datetime |
|
|
| import gradio as gr |
| import numpy as np |
| import requests |
| from PIL import Image, ImageDraw, ImageFont |
| from ultralytics import YOLO |
|
|
| DETECTOR_MODEL = "yolov8n.pt" |
| MAX_OBJECTS = 12 |
| MODAL_EXCAVATE_URL = os.getenv("MODAL_EXCAVATE_URL", "").strip() |
|
|
| _detector = None |
|
|
|
|
| PERIODS = [ |
| ("Pleistocene of Productivity", "pre-2016", "fossilized adapters, legacy paper deposits"), |
| ("Early USB-C Transition", "2017-2019", "mixed connector ecologies and dongle migration"), |
| ("Pandemic Desk Accretion", "2020-2022", "snack bowls, webcam altars, duplicated notebooks"), |
| ("Late Notification Age", "2023-present", "charging cables, earbuds, hydration vessels"), |
| ] |
|
|
|
|
| def get_detector(): |
| global _detector |
| if _detector is None: |
| _detector = YOLO(DETECTOR_MODEL) |
| return _detector |
|
|
|
|
| def detect_artifacts(image: Image.Image): |
| model = get_detector() |
| result = model.predict(image, imgsz=960, conf=0.18, verbose=False)[0] |
| artifacts = [] |
| width, height = image.size |
| for box in result.boxes[:MAX_OBJECTS]: |
| x1, y1, x2, y2 = [float(v) for v in box.xyxy[0].tolist()] |
| cls_id = int(box.cls[0]) |
| label = model.names.get(cls_id, f"object_{cls_id}") |
| conf = float(box.conf[0]) |
| area = ((x2 - x1) * (y2 - y1)) / max(width * height, 1) |
| cx = (x1 + x2) / 2 / width |
| cy = (y1 + y2) / 2 / height |
| artifacts.append( |
| { |
| "id": len(artifacts) + 1, |
| "label": label, |
| "confidence": round(conf, 2), |
| "box": [x1, y1, x2, y2], |
| "center": [round(cx, 3), round(cy, 3)], |
| "area": round(area, 4), |
| "period": random.choice(PERIODS), |
| } |
| ) |
| if not artifacts: |
| artifacts.append( |
| { |
| "id": 1, |
| "label": "unclassified clutter horizon", |
| "confidence": 0.42, |
| "box": [width * 0.2, height * 0.2, width * 0.8, height * 0.8], |
| "center": [0.5, 0.5], |
| "area": 0.36, |
| "period": random.choice(PERIODS), |
| } |
| ) |
| return artifacts |
|
|
|
|
| def annotate_image(image: Image.Image, artifacts): |
| annotated = image.convert("RGB").copy() |
| draw = ImageDraw.Draw(annotated, "RGBA") |
| try: |
| font = ImageFont.truetype("DejaVuSans-Bold.ttf", max(18, image.width // 40)) |
| small = ImageFont.truetype("DejaVuSans.ttf", max(12, image.width // 70)) |
| except Exception: |
| font = small = ImageFont.load_default() |
|
|
| colors = [(255, 80, 80), (60, 180, 255), (255, 190, 60), (140, 255, 120), (210, 120, 255)] |
| for artifact in artifacts: |
| x1, y1, x2, y2 = artifact["box"] |
| color = colors[(artifact["id"] - 1) % len(colors)] |
| draw.rectangle([x1, y1, x2, y2], outline=(*color, 240), width=max(3, image.width // 250)) |
| r = max(14, image.width // 55) |
| cx, cy = x1 + r + 4, y1 + r + 4 |
| draw.ellipse([cx - r, cy - r, cx + r, cy + r], fill=(*color, 230), outline=(0, 0, 0, 255), width=2) |
| label = str(artifact["id"]) |
| bbox = draw.textbbox((0, 0), label, font=font) |
| draw.text((cx - (bbox[2] - bbox[0]) / 2, cy - (bbox[3] - bbox[1]) / 2 - 1), label, fill=(0, 0, 0), font=font) |
| draw.text((x1, max(0, y1 - 18)), artifact["label"], fill=(255, 255, 255), font=small, stroke_width=2, stroke_fill=(0, 0, 0)) |
| return annotated |
|
|
|
|
| def make_site_map(artifacts): |
| rows = ["| Find | Artifact | Grid position | Confidence |", "|---:|---|---|---:|"] |
| for a in artifacts: |
| cx, cy = a["center"] |
| grid = f"{chr(65 + min(4, int(cx * 5)))}-{1 + min(4, int(cy * 5))}" |
| rows.append(f"| {a['id']} | {a['label']} | {grid} | {a['confidence']:.2f} |") |
| return "\n".join(rows) |
|
|
|
|
| def fallback_report(artifacts, tone): |
| counts = Counter(a["label"] for a in artifacts) |
| lead = counts.most_common(1)[0][0] |
| lines = [ |
| "# Preliminary Archaeological Report: Domestic Sector A", |
| f"**Excavation date:** {datetime.utcnow().strftime('%Y-%m-%d UTC')} ", |
| f"**Interpretive mode:** {tone}", |
| "", |
| "## Abstract", |
| f"Survey of the submitted room photograph identified {len(artifacts)} loci of material culture. The dominant visible taxon, **{lead}**, suggests a civilization balancing productivity with ceremonial postponement.", |
| "", |
| "## Artifact catalog", |
| ] |
| for a in artifacts: |
| period, date_range, clue = a["period"] |
| cx, cy = a["center"] |
| placement = "peripheral" if cx < 0.25 or cx > 0.75 else "central" |
| elevation = "upper stratum" if cy < 0.45 else "lower stratum" |
| lines.append( |
| f"- **Clue #{a['id']}: {a['label'].title()}** — Recovered from the {placement} {elevation}. " |
| f"Carbon-ish dating assigns it to the **{period}** ({date_range}), based on {clue}. " |
| f"Its placement indicates ritual readiness, abandonment, or a tiny rebellion against storage furniture." |
| ) |
| lines.extend( |
| [ |
| "", |
| "## Spatial interpretation", |
| "Objects form a loose constellation around probable human activity zones. This pattern is consistent with the well-known 'I'll move it later' deposition model (Smith, 1999; Patel & Cordova, 2018).", |
| "", |
| "## Conclusion", |
| f"The site is stable but culturally active. Recommended next step: preserve the chaos for peer review, unless guests are arriving within 45 minutes.", |
| ] |
| ) |
| return "\n".join(lines) |
|
|
|
|
| def modal_report(artifacts, tone): |
| if not MODAL_EXCAVATE_URL: |
| return None |
| try: |
| payload = {"artifacts": artifacts, "tone": tone} |
| response = requests.post(MODAL_EXCAVATE_URL, json=payload, timeout=90) |
| response.raise_for_status() |
| return response.json().get("report") |
| except Exception as exc: |
| return f"_Modal report generation failed, using local curator notes instead: {exc}_\n\n" + fallback_report(artifacts, tone) |
|
|
|
|
| def timeline_html(artifacts): |
| grouped = defaultdict(list) |
| for a in artifacts: |
| grouped[a["period"][0]].append(a) |
| cards = [] |
| for period, date_range, clue in PERIODS: |
| items = grouped.get(period, []) |
| artifact_text = ", ".join(f"#{a['id']} {a['label']}" for a in items) or "no visible deposits" |
| cards.append( |
| f"<div class='era'><b>{period}</b><span>{date_range}</span><p>{artifact_text}</p><small>{clue}</small></div>" |
| ) |
| return "<div class='timeline'>" + "".join(cards) + "</div>" |
|
|
|
|
| def excavate(image, tone): |
| if image is None: |
| raise gr.Error("Upload a room, desk, or workspace photo before excavating.") |
| if not isinstance(image, Image.Image): |
| image = Image.fromarray(np.array(image)) |
| artifacts = detect_artifacts(image.convert("RGB")) |
| annotated = annotate_image(image, artifacts) |
| report = modal_report(artifacts, tone) or fallback_report(artifacts, tone) |
| return annotated, make_site_map(artifacts), report, timeline_html(artifacts), json.dumps(artifacts, indent=2) |
|
|
|
|
| CSS = """ |
| #title {text-align:center} .timeline{display:grid;grid-template-columns:repeat(auto-fit,minmax(180px,1fr));gap:12px}.era{border:1px solid #ddd;border-radius:14px;padding:14px;background:linear-gradient(180deg,#fff8e7,#f5efe3)}.era span{display:block;color:#7a4b12;font-size:.9rem;margin:.25rem 0}.era p{min-height:3rem}.era small{color:#665} |
| """ |
|
|
| with gr.Blocks(title="Photo Archaeology Simulator", theme=gr.themes.Soft(), css=CSS) as demo: |
| gr.Markdown( |
| """ |
| # Photo Archaeology Simulator 🏺 |
| |
| Turn your messy room into an academic archaeological dig site report. Upload a chaotic room/workspace photo, then let the excavation team number artifacts, draft a site map, and fabricate scholarly domestic stratigraphy. |
| """, |
| elem_id="title", |
| ) |
| with gr.Row(): |
| with gr.Column(scale=1): |
| image = gr.Image(type="pil", label="Chaotic site photo") |
| tone = gr.Radio(["Very academic", "Maximum nonsense", "Museum placard", "Grant proposal"], value="Very academic", label="Report style") |
| btn = gr.Button("Excavate", variant="primary") |
| with gr.Column(scale=1): |
| annotated = gr.Image(type="pil", label="Annotated site map") |
| with gr.Row(): |
| site_map = gr.Markdown(label="Numbered locations") |
| report = gr.Markdown(label="Scholarly report") |
| timeline = gr.HTML(label="Timeline") |
| raw = gr.Code(label="Artifact JSON", language="json", visible=False) |
| btn.click(excavate, [image, tone], [annotated, site_map, report, timeline, raw]) |
|
|
| if __name__ == "__main__": |
| demo.launch() |
|
|
