app.py

import os
import cv2
import gradio as gr
import google.generativeai as genai
from ultralytics import YOLO
import tempfile
import torch
import spaces

import numpy as np
from PIL import Image, ImageDraw, ImageFont
import arabic_reshaper
from bidi.algorithm import get_display

# =============================
# Gemini API Key
# =============================
# ⚠️ الصق مفتاحك محليًا هنا داخل ملفك (لا تنشره بمستودع عام)
GEMINI_API_KEY = "AIzaSyAvm28ZnTMaZ1Jtg9sYM-EO4qlAN2W4BIQ"

# خيار "أقل خطورة": لو موجود Secrets/Env استخدمه بدل المكتوب
# GEMINI_API_KEY = os.getenv("GEMINI_API_KEY") or "PASTE_YOUR_GEMINI_KEY_HERE"

genai.configure(api_key=GEMINI_API_KEY)

SYSTEM_PROMPT = (
    "لدي نص خام عبارة عن حروف عربية متتابعة بدون مسافات "
    "ومع وجود تكرار بسيط لأنه ناتج من مترجم لغة الإشارة.\n"
    "مهمتك:\n"
    "1) إزالة التكرار غير الضروري.\n"
    "2) إضافة المسافات بين الكلمات.\n"
    "3) إخراج الجملة الأقرب للمعنى.\n"
    "أعد النص فقط بدون شرح."
)

def fix_with_gemini(raw_text: str) -> str:
    if not raw_text:
        return ""
    try:
        model = genai.GenerativeModel("models/gemini-2.5-flash")
        prompt = SYSTEM_PROMPT + f"\n\nالنص الخام:\n«{raw_text}»"
        resp = model.generate_content(prompt)
        return (resp.text or "").strip()
    except Exception as e:
        return f"خطأ في Gemini: {e}"


# =============================
# إعدادات YOLO
# =============================
WEIGHTS_PATH = "best.pt"
IMG_SIZE = 1080
CONF_THRESHOLD = 0.15

MIN_STABLE_FRAMES = 1
FRAME_SKIP = 1
MAX_FRAMES = 1000
WORD_GAP_FRAMES = 10

CENTER_CROP = True

arabic_map = {
    "aleff": "ا",
    "bb": "ب",
    "ta": "ت",
    "taa": "ت",
    "thaa": "ث",
    "jeem": "ج",
    "haa": "ح",
    "khaa": "خ",
    "dal": "د",
    "dha": "ظ",
    "dhad": "ض",
    "fa": "ف",
    "gaaf": "ق",
    "ghain": "غ",
    "ha": "ه",
    "kaaf": "ك",
    "laam": "ل",
    "meem": "م",
    "nun": "ن",
    "ra": "ر",
    "saad": "ص",
    "seen": "س",
    "sheen": "ش",
    "thal": "ذ",
    "toot": "ة",
    "waw": "و",
    "ya": "ي",
    "yaa": "ي",
    "zay": "ز",
    "ain": "ع",
    "al": "ال",
    "la": "لا",
}

yolo_model = None
DEVICE = "cpu"

def get_model():
    global yolo_model, DEVICE

    if yolo_model is None:
        print("🔹 Loading YOLO model...")
        yolo_model = YOLO(WEIGHTS_PATH)
        print("📚 Classes:", yolo_model.names)

    if torch.cuda.is_available():
        if DEVICE != "cuda":
            DEVICE = "cuda"
            try:
                yolo_model.to(DEVICE)
                print("✅ YOLO model moved to cuda")
            except Exception as e:
                print("⚠️ تعذر نقل الموديل إلى cuda:", e)
    else:
        if DEVICE != "cpu":
            print("⚠️ CUDA غير متوفر، سيتم استخدام CPU.")
        DEVICE = "cpu"

    return yolo_model


# =============================
# إصلاح ????: رسم عربي على الفيديو via PIL
# =============================
FONT_PATH = os.path.join(os.path.dirname(__file__), "NotoNaskhArabic-VariableFont_wght.ttf")

def draw_arabic_text(frame_bgr, text, x, y, font_size=36, bgr_color=(0, 255, 0)):
    img = Image.fromarray(cv2.cvtColor(frame_bgr, cv2.COLOR_BGR2RGB))
    draw = ImageDraw.Draw(img)

    try:
        font = ImageFont.truetype(FONT_PATH, font_size)
    except Exception as e:
        print("⚠️ خطأ تحميل الخط العربي:", e)
        font = ImageFont.load_default()

    shaped = arabic_reshaper.reshape(text)
    rtl_text = get_display(shaped)

    rgb_color = (bgr_color[2], bgr_color[1], bgr_color[0])
    draw.text((x, y), rtl_text, font=font, fill=rgb_color)

    return cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)


# =============================
# تكبير + قص من الوسط 640x640
# =============================
def resize_and_center_crop(frame, target: int = 640):
    h, w = frame.shape[:2]
    short_side = min(w, h)
    if short_side <= 0:
        return frame

    scale = target / short_side
    new_w = int(w * scale)
    new_h = int(h * scale)

    frame = cv2.resize(frame, (new_w, new_h), interpolation=cv2.INTER_AREA)

    h, w = frame.shape[:2]
    x1 = max(0, (w - target) // 2)
    y1 = max(0, (h - target) // 2)
    x2 = min(x1 + target, w)
    y2 = min(y1 + target, h)

    crop = frame[y1:y2, x1:x2]

    ch, cw = crop.shape[:2]
    if ch != target or cw != target:
        crop = cv2.resize(crop, (target, target), interpolation=cv2.INTER_AREA)

    return crop


# =============================
# تجهيز الفيديو قبل المعالجة
# =============================
def preprocess_video(input_path: str, target_short_side: int = 1080, target_fps: int = 8) -> str:
    cap = cv2.VideoCapture(input_path)
    if not cap.isOpened():
        print("[preprocess] تعذر فتح الفيديو، سنستخدم الملف الأصلي كما هو.")
        return input_path

    orig_fps = cap.get(cv2.CAP_PROP_FPS)
    w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

    if orig_fps <= 0:
        frame_step = 1
        out_fps = float(target_fps)
    else:
        frame_step = max(1, int(round(orig_fps / target_fps)))
        out_fps = orig_fps / frame_step

    short_side = min(w, h)
    scale = 1.0 if short_side <= 0 else (target_short_side / short_side)
    new_w = int(w * scale)
    new_h = int(h * scale)

    fd, tmp_path = tempfile.mkstemp(suffix=".mp4")
    os.close(fd)

    out_w, out_h = (IMG_SIZE, IMG_SIZE) if CENTER_CROP else (new_w, new_h)

    fourcc = cv2.VideoWriter_fourcc(*"mp4v")
    out = cv2.VideoWriter(tmp_path, fourcc, out_fps, (out_w, out_h))

    frame_idx = 0
    while True:
        ret, frame = cap.read()
        if not ret:
            break

        if frame_idx % frame_step == 0:
            if CENTER_CROP:
                processed = resize_and_center_crop(frame, target=IMG_SIZE)
            else:
                processed = cv2.resize(frame, (new_w, new_h), interpolation=cv2.INTER_AREA)
            out.write(processed)

        frame_idx += 1

    cap.release()
    out.release()
    print(f"[preprocess] orig=({w}x{h}), new=({out_w}x{out_h}), saved={tmp_path}")
    return tmp_path


# =============================
# معالجة فريم واحد
# =============================
def detect_frame(frame_bgr):
    model = get_model()

    frame_rgb = cv2.cvtColor(frame_bgr, cv2.COLOR_BGR2RGB)
    result = model.predict(
        frame_rgb,
        conf=CONF_THRESHOLD,
        imgsz=IMG_SIZE,
        verbose=False,
        device=DEVICE,
    )[0]

    boxes = result.boxes
    num_boxes = 0 if boxes is None else len(boxes)
    print(f"[detect_frame] boxes={num_boxes}")

    if boxes is None or len(boxes) == 0:
        return [], frame_bgr

    labels = []
    for box in boxes:
        x1, y1, x2, y2 = map(int, box.xyxy[0])
        cls_id = int(box.cls[0])

        if isinstance(model.names, dict):
            eng = model.names.get(cls_id, str(cls_id))
        else:
            eng = model.names[cls_id] if cls_id < len(model.names) else str(cls_id)

        letter = arabic_map.get(eng, eng)
        labels.append(letter)

        cv2.rectangle(frame_bgr, (x1, y1), (x2, y2), (0, 255, 0), 2)
        frame_bgr = draw_arabic_text(frame_bgr, letter, x1, max(0, y1 - 45), font_size=36)

    return labels, frame_bgr


# =============================
# VIDEO → RAW TEXT + OUTPUT VIDEO + DEBUG
# =============================
def extract_and_render(video_path: str):
    cap = cv2.VideoCapture(video_path)
    if not cap.isOpened():
        return "", None, "تعذر فتح الفيديو في extract_and_render"

    fourcc = cv2.VideoWriter_fourcc(*"mp4v")
    out_path = "processed_output.mp4"

    fps = cap.get(cv2.CAP_PROP_FPS)
    width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
    if fps <= 0:
        fps = 8.0

    out = cv2.VideoWriter(out_path, fourcc, fps, (width, height))

    word = ""
    words = []
    last_label = None
    last_added = None
    stable = 0
    last_seen = None
    frame_index = 0

    frames_with_dets = 0
    debug_lines = []

    while True:
        ret, frame = cap.read()
        if not ret:
            break

        frame_index += 1
        if frame_index > MAX_FRAMES:
            break

        if FRAME_SKIP > 1 and frame_index % FRAME_SKIP != 0:
            continue

        frame = cv2.flip(frame, 1)
        labels, rendered = detect_frame(frame)
        out.write(rendered)

        if labels:
            frames_with_dets += 1
            debug_lines.append(f"frame {frame_index}: {labels}")

            label = labels[0]
            last_seen = frame_index

            if label == last_label:
                stable += 1
            else:
                last_label = label
                stable = 1

            if stable >= MIN_STABLE_FRAMES:
                if label != last_added:
                    word += label
                    last_added = label
                stable = 0
        else:
            if word and last_seen and (frame_index - last_seen >= WORD_GAP_FRAMES):
                words.append(word)
                word = ""
                last_label = None
                last_added = None
                stable = 0
                last_seen = None

    cap.release()
    out.release()

    if word:
        words.append(word)

    raw_text = " ".join(words).strip()

    if not debug_lines:
        debug_info = (
            f"total_frames={frame_index}, frames_with_detections=0\n"
            "لم يتم رصد أي صناديق (boxes) من YOLO في أي فريم.\n"
            "تحقق من:\n"
            "- أن best.pt هو موديل detection وتدريبه سليم.\n"
            "- أن الفيديو مشابه لتدريب الموديل من ناحية وضعية اليد والكاميرا."
        )
    else:
        sample = "\n".join(debug_lines[:30])
        debug_info = (
            f"total_frames={frame_index}, frames_with_detections={frames_with_dets}\n"
            "أمثلة من الفريمات اللي فيها حروف:\n"
            f"{sample}"
        )

    return raw_text, out_path, debug_info


# =============================
# Gradio + @spaces.GPU
# =============================
@spaces.GPU
def run(file):
    if file is None:
        return "لم يتم رفع فيديو", "", None, "لم يتم رفع فيديو"

    video_path = file.name
    light_path = preprocess_video(video_path, target_short_side=640, target_fps=8)

    raw, processed_path, debug_info = extract_and_render(light_path)
    pretty = fix_with_gemini(raw) if raw else ""

    if not raw:
        raw = "لم يتم التعرف على أي نص من الإشارات."

    return raw, pretty, processed_path, debug_info


with gr.Blocks() as demo:
    gr.Markdown("## 🤟 ASL → Arabic (YOLO + Gemini) — إصلاح ظهور الحروف العربية داخل الفيديو")

    inp = gr.File(label="ارفع فيديو الإشارة")
    raw = gr.Textbox(label="النص الخام", lines=3)
    pretty = gr.Textbox(label="النص المحسن (Gemini)", lines=3)
    video_out = gr.Video(label="الفيديو بعد البروسيس")
    debug_box = gr.Textbox(label="Debug info", lines=10)

    btn = gr.Button("ابدأ المعالجة")
    btn.click(run, inputs=[inp], outputs=[raw, pretty, video_out, debug_box])

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)