move face model

backend/face_model/download_model_image.py

@@ -0,0 +1,58 @@
+import os
+import requests
+from bs4 import BeautifulSoup
+import cv2
+import numpy as np
+
+
+def download_images(search_query, num_images, output_dir):
+    url = f"https://www.bing.com/images/search?q={search_query}&form=HDRSC2"
+    headers = {
+        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/64.0.3282.140 Safari/537.36 Edge/17.17134"
+    }
+
+    response = requests.get(url, headers=headers)
+    soup = BeautifulSoup(response.text, "html.parser")
+    image_tags = soup.find_all("img", {"class": "mimg"})
+
+    if not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+
+    i = 0
+    face_cascade = cv2.CascadeClassifier(
+        cv2.data.haarcascades + "haarcascade_frontalface_default.xml"
+    )
+
+    for img_tag in image_tags:
+        img_url = img_tag.get("src") or img_tag.get("data-src")
+
+        # Ignore data URL
+        if img_url.startswith("data:image"):
+            continue
+
+        img_data = requests.get(img_url).content
+        img = np.asarray(bytearray(img_data), dtype="uint8")
+        img = cv2.imdecode(img, cv2.IMREAD_COLOR)
+        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+        faces = face_cascade.detectMultiScale(
+            gray, scaleFactor=1.05, minNeighbors=5, minSize=(40, 40)
+        )
+
+        if len(faces) == 1:
+            with open(os.path.join(output_dir, f"{search_query}_{i}.jpg"), "wb") as f:
+                f.write(img_data)
+            print(f"Downloaded {i + 1}/{num_images} images")
+
+            i += 1
+            if i >= num_images:
+                break
+
+
+if __name__ == "__main__":
+    politician_names = ["伊藤博文"]
+    num_images = 10
+    base_output_dir = "./politicians"
+
+    for name in politician_names:
+        output_dir = os.path.join(base_output_dir, name)
+        download_images(name, num_images, output_dir)

backend/package.json → backend/face_model/package.json

@@ -1,7 +1,6 @@
 {
-  "name": "politicians",
+  "name": "face_model",
   "version": "1.0.0",
-  "description": "顔認識用の画像データ",
   "main": "index.js",
   "scripts": {
     "test": "echo \"Error: no test specified\" && exit 1"

backend/lib/face/face_icon_extractor.py

@@ -0,0 +1,43 @@
+import cv2
+import urllib.request
+import numpy as np
+
+
+def url_to_image(url):
+    with urllib.request.urlopen(url) as resp:
+        image = np.asarray(bytearray(resp.read()), dtype="uint8")
+        image = cv2.imdecode(image, cv2.IMREAD_COLOR)
+    return image
+
+
+def extract_face_icon_from_url(image_url, casc_path="haarcascade_frontalface_default.xml", padding_ratio=0.7):
+    face_cascade = cv2.CascadeClassifier(casc_path)
+
+    image = url_to_image(image_url)
+    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+
+    faces = face_cascade.detectMultiScale(
+        gray, scaleFactor=1.1, minNeighbors=5, minSize=(30, 30))
+
+    if len(faces) == 0:
+        return None
+
+    x, y, w, h = faces[0]
+    center_x = x + w // 2
+    center_y = y + h // 2
+
+    square_length = min(w, h) * (1 + padding_ratio)
+
+    start_x = int(center_x - square_length // 2)
+    start_y = int(center_y - square_length // 2)
+    end_x = int(start_x + square_length)
+    end_y = int(start_y + square_length)
+
+    start_x = max(0, start_x)
+    start_y = max(0, start_y)
+    end_x = min(image.shape[1], end_x)
+    end_y = min(image.shape[0], end_y)
+
+    icon = image[start_y:end_y, start_x:end_x]
+
+    return icon