from PIL import Image
import numpy as np
import gradio as gr
import os
import cv2
from nudenet import NudeDetector
import io
import concurrent.futures

# ─── Konstanten ────────────────────────────────────────
DETECTION_MAX_DIM    = 768
PIXELS_PER_CM_ESTIMATE = 15
MIN_CONFIDENCE       = 0.45

# ─── Hilfsfunktionen ───────────────────────────────────
def resize_for_detection(img_pil, max_dim):
    if max(img_pil.width, img_pil.height) <= max_dim:
        return img_pil, 1.0
    ratio = max_dim / max(img_pil.width, img_pil.height)
    new_size = (int(img_pil.width * ratio), int(img_pil.height * ratio))
    resized = img_pil.resize(new_size, Image.Resampling.LANCZOS)
    return resized, 1 / ratio

# Deine bestehenden describe-Funktionen (unverändert)
def describe_breast_precise(crop_pil): 
    # ... dein Code bleibt gleich ...
    return "Form: Rund · Größe: mittel · Nippel: Sichtbar · 9.8×8.4 cm"   # Beispiel

def describe_vagina_precise(crop_pil): 
    # ... dein Code bleibt gleich ...
    return "Form: Klassisches Outie · Größe: mittel · Prominenz: sichtbar · Behaart: minimal · 8.2×11.1 cm"

detector = NudeDetector(inference_resolution=640)

def process_single_image(mode: str, path: str):
    try:
        original_pil = Image.open(path).convert("RGB")
        detection_pil, scale = resize_for_detection(original_pil, DETECTION_MAX_DIM)
        
        detections = detector.detect(np.array(detection_pil))
        
        target_class = "FEMALE_BREAST_EXPOSED" if mode == "Brüste" else "FEMALE_GENITALIA_EXPOSED"
        relevant = [d for d in detections if d["class"] == target_class and d.get("score", 0) >= MIN_CONFIDENCE]
        
        filename = os.path.basename(path)
        markdown = f"**{filename}** — {mode}\n\n"
        
        outputs = []
        
        if not relevant:
            markdown += "❌ Keine relevanten Bereiche erkannt.\n"
            outputs.append((None, markdown))
            return outputs
        
        markdown += f"✅ **{len(relevant)}** {mode.lower()} gefunden\n\n"
        
        for i, det in enumerate(relevant, 1):
            x, y, w, h = [int(v * scale) for v in det["box"]]
            crop_pil = original_pil.crop((x, y, x + w, y + h))
            
            desc = describe_breast_precise(crop_pil) if mode == "Brüste" else describe_vagina_precise(crop_pil)
            
            # Crop als bytes für Gradio Image-Komponente
            crop_bytes = io.BytesIO()
            crop_pil.save(crop_bytes, format="PNG")
            crop_bytes.seek(0)
            
            markdown_part = f"**{mode} {i}**  (Konfidenz: {det['score']:.2f})\n{desc}\n"
            
            outputs.append((crop_bytes, markdown_part))
        
        # Zusammenfassung am Ende
        summary_md = "\n".join([md for _, md in outputs]) + f"\n\n**Gesamt: {len(relevant)} Funde**"
        outputs.append((None, summary_md))
        
        return outputs
    
    except Exception as e:
        return [(None, f"**{filename}** — Fehler: {str(e)}")]

def analyze_images(mode: str, image_paths):
    if not image_paths:
        return [(None, "**Keine Bilder hochgeladen.**")]
    
    all_outputs = []
    
    max_workers = min(6, len(image_paths), os.cpu_count() or 4)
    with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
        futures = [executor.submit(process_single_image, mode, p) for p in image_paths]
        for future in concurrent.futures.as_completed(futures):
            all_outputs.extend(future.result())
    
    return all_outputs

# ─── Gradio Interface ──────────────────────────────────
with gr.Blocks(theme=gr.themes.Soft(primary_hue="pink", secondary_hue="purple")) as demo:
    gr.Markdown("# Nackt-Analyzer – mit Crops")
    gr.Markdown("Lädt Bilder → erkennt Brüste / Vulva → zeigt **Text + Crop-Bilder**")
    
    with gr.Row():
        mode = gr.Radio(choices=["Brüste", "Vagina"], value="Brüste", label="Modus")
        upload = gr.File(file_count="multiple", file_types=["image"], label="Bilder hochladen")
    
    analyze_btn = gr.Button("Analysieren", variant="primary")
    
    output_gallery = gr.Gallery(
        label="Ergebnisse (Crops + Beschreibung)",
        columns=3,
        height="auto",
        object_fit="contain",
        show_label=True,
        elem_id="result-gallery"
    )
    
    markdown_output = gr.Markdown(label="Zusammenfassung / Details")
    
    def on_analyze(mode, files):
        if not files:
            return [], "**Keine Dateien ausgewählt.**"
        
        paths = [f.name for f in files] if hasattr(files[0], 'name') else files
        results = analyze_images(mode, paths)
        
        images = []
        md_parts = []
        
        for img_bytes, text in results:
            if img_bytes is not None:
                images.append((img_bytes, text))   # (image, caption)
            else:
                md_parts.append(text)
        
        combined_md = "\n\n".join(md_parts) if md_parts else ""
        
        return images, combined_md
    
    analyze_btn.click(
        fn=on_analyze,
        inputs=[mode, upload],
        outputs=[output_gallery, markdown_output]
    )

if __name__ == "__main__":
    demo.launch(share=True)