Update demo pipeline

README.md

@@ -1,12 +1,10 @@
 ---
 title: MNEMO OCR Demo
 emoji: 🔍
-colorFrom: purple
-colorTo: gray
 sdk: gradio
 sdk_version: 3.41.2
 app_file: app.py
 pinned: false
 ---
 
-MNEMO OCR – simple OCR demo using PaddleOCR.
+Production-like demo of sensor digitization pipeline.

app.py

@@ -1,41 +1,73 @@
 import gradio as gr
-from paddleocr import PaddleOCR
-from PIL import Image
 import numpy as np
+import cv2
+import pandas as pd
+from pathlib import Path
+from tempfile import NamedTemporaryFile
 
-# Инициализируем OCR один раз
-ocr = PaddleOCR(
-    use_angle_cls=True,
-    lang='en'  # если нужно rus — скажи
-)
+from src.pipeline_hf import process_single_image
+from src.config_loader import load_config
+
+
+# ================================
+# Вспомогательная функция: подсветка сенсоров
+# ================================
+def draw_boxes(img_rgb, sensors):
+    img = img_rgb.copy()
+
+    for s in sensors:
+        x, y, w, h = s["x"], s["y"], s["w"], s["h"]
+        cv2.rectangle(img, (x, y), (x + w, y + h), (0, 255, 0), 2)
+
+    return img
+
+
+# ================================
+# Основная функция демо
+# ================================
+def hf_process(img_rgb):
+
+    cfg = load_config("configs/config.yaml")
+
+    # Запуск пайплайна
+    result = process_single_image(img_rgb, cfg_path="configs/config.yaml")
+
+    title = result["title"]
+    sensors = result["sensors"]
 
-def run_ocr(image):
-    if image is None:
-        return "No image uploaded."
+    # --- визуализация сенсоров ---
+    boxed = draw_boxes(img_rgb, sensors)
 
-    # Конвертируем PIL → numpy
-    img = np.array(image)
+    # --- DataFrame сенсоров ---
+    df = pd.DataFrame(sensors)[["text", "score", "x", "y", "w", "h"]]
 
-    # Запуск OCR
-    result = ocr.ocr(img)
+    # --- Excel на скачивание ---
+    tmp = NamedTemporaryFile(delete=False, suffix=".xlsx")
+    df.to_excel(tmp.name, index=False)
 
-    # Формируем вывод текстом
-    lines = []
-    for block in result:
-        for line in block:
-            text = line[1][0]
-            conf = line[1][1]
-            lines.append(f"{text} (conf: {conf:.2f})")
+    return (
+        boxed,
+        title,
+        df,
+        tmp.name  # путь к скачиваемому файлу
+    )
 
-    return "\n".join(lines)
 
+# ================================
+# Gradio UI
+# ================================
 demo = gr.Interface(
-    fn=run_ocr,
-    inputs=gr.Image(type="pil"),
-    outputs=gr.Textbox(label="Recognized Text"),
-    title="MNEMO OCR Demo",
-    description="Upload an image and extract text using PaddleOCR.",
+    fn=hf_process,
+    inputs=gr.Image(type="numpy", label="Upload image"),
+    outputs=[
+        gr.Image(label="Detected sensors"),
+        gr.Textbox(label="Title"),
+        gr.Dataframe(label="Recognized sensors"),
+        gr.File(label="Download Excel")
+    ],
+    title="MNEMO OCR — HF Demo",
+    description="Продовая демо-версия оцифровщика MNEMO."
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()

requirements.txt

@@ -1,5 +1,11 @@
-gradio==3.41.2
+numpy==1.23.5
+pandas==2.2.3
+pillow==10.4.0
+opencv-python-headless==4.7.0.72
+
 paddlepaddle==2.6.1
 paddleocr==2.6.1.0
-numpy==1.23.5
-pillow==10.4.0
+
+gradio==3.41.2
+pyyaml==6.0.2
+openpyxl==3.1.5

src/pipeline_hf.py

@@ -0,0 +1,131 @@
+import cv2
+import numpy as np
+import pandas as pd
+
+from pathlib import Path
+from src.config_loader import load_config
+from src.ocr_utils_demo import ocr_title, ocr_sensors
+
+def process_single_image(img: np.ndarray, cfg_path: str | Path = "configs/config.yaml"):
+    """
+    Обрабатывает одно изображение:
+    - вытаскивает титул;
+    - находит и оцифровывает сенсоры;
+    - возвращает структуру с результатом.
+    """
+    cfg = load_config(cfg_path)
+
+    # Цветовые диапазоны сенсоров
+    color_ranges = []
+
+    for key, rng in color_ranges.items():
+        lo = np.array(rng["from"], dtype=np.uint8)
+        hi = np.array(rng["to"], dtype=np.uint8)
+        color_ranges.append((lo, hi))
+
+    # ---------- 1. Оцифровка титула ----------
+    h, w = img.shape[:2]
+
+    img_bgr = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+
+    title_roi = img_bgr[:45, :int(w / 2.4)]
+    title_text = ocr_title(title_roi)
+
+    # ---------- 2. Оцифровка сенсоров ----------
+    hsv = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2HSV)
+
+    mask = None
+
+    for lo_np, hi_np in color_ranges:
+        cur = cv2.inRange(hsv, lo_np, hi_np)
+        mask = cur if mask is None else cv2.bitwise_or(mask, cur)
+
+    contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+
+    rois, positions = [], []
+    for cnt in contours:
+        x, y, ww, hh = cv2.boundingRect(cnt)
+        if ww < 90 or hh < 17:
+            continue
+
+        hh_clamped = min(hh, 17)
+
+        roi = img_bgr[y:y + hh_clamped, x:x + ww]
+
+        rois.append(roi)
+        positions.append((x, y, ww, hh_clamped))
+
+    # ---------- 3. OCR сенсоров ----------
+    sensors = []
+    if rois:
+        ocr_results = ocr_sensors(rois)
+
+        for (x, y, ww, hh), r in zip(positions, ocr_results):
+            sensors.append({
+                "text": r["text"],
+                "score": r["score"],
+                "x": x,
+                "y": y,
+                "w": ww,
+                "h": hh
+            })
+
+    return {
+        "title": title_text,
+        "sensors": sensors
+    }
+
+
+# ---------------------------------------------------------
+# Основной pipeline → Excel
+# ---------------------------------------------------------
+def process_image_to_excel(cfg: dict):
+    input_dir = Path(cfg["paths"]["input"])
+    output_dir = Path(cfg["paths"]["output"])
+    excel_name = cfg["export"]["excel_filename"]
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+    excel_path = output_dir / excel_name
+
+    results = []
+
+    for img_path in input_dir.glob("*.png"):
+        name = img_path.name
+
+        img = cv2.imread(str(img_path))
+        if img is None:
+            print(f"Не могу прочитать файл {name}")
+            continue
+
+        img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+
+        res = process_single_image(img_rgb, cfg_path="configs/config.yaml")
+        
+        title_text = res["title"]
+        sensors = res["sensors"]
+
+        for sen in sensors:
+            results.append({
+                "filename": name,
+                "title": title_text,
+                "sensor_name": sen["text"],
+                "score": sen["score"]
+            })
+
+    if not results:
+        print("⚠ Нет данных для записи.")
+        return
+
+    df = pd.DataFrame(results)
+    df.to_excel(excel_path, index=False, engine="openpyxl")
+
+    print(f"✅ Готово! Excel сохранён: {excel_path.resolve()}")
+
+
+# ---------------------------------------------------------
+# Запуск
+# ---------------------------------------------------------
+if __name__ == "__main__":
+    cfg = load_config("configs/config.yaml")
+    process_image_to_excel(cfg)
+