get image info

requirements.txt

 faiss-cpu==1.6.3
-pandas==1.1.2
+pandas==1.1.1
 elasticsearch==7.9.1
 opencv-python==4.4.0.42
 pillow==7.2.0

src/diary_cover_similarity.py

+import os
 import time
 
-from utils.images import (BASE_DIR, MODELS_DIR, face_to_vec, file_to_ndarray, url_to_ndarray)
+import dlib
+
+from utils.images import face_to_vec, file_to_ndarray, url_to_ndarray
 
 
 def main():
@@ -8,21 +11,30 @@ def main():
     img = url_to_ndarray(img_url)
     print(img)
 
-    faces = face_to_vec(img)
+    base_dir = os.getcwd()
+    print("base_dir: " + base_dir)
+    model_diry = os.path.join(base_dir, "_models")
+    facerec_model_path = os.path.join(model_diry, "dlib_face_recognition_resnet_model_v1.dat")
+    shape_model_path = os.path.join(model_diry, "shape_predictor_68_face_landmarks.dat")
+
+    face_rec = dlib.face_recognition_model_v1(facerec_model_path)
+    face_detector = dlib.get_frontal_face_detector()
+    shape_predictor = dlib.shape_predictor(shape_model_path)
+
+    faces = face_to_vec(img, face_rec, face_detector, shape_predictor)
 
     print(faces)
     print("============")
 
     for face in faces:
-        print(face["feature"] + "\n")
+        print(face)
+        print("\n")
+        print("-------------")
 
 
 if __name__ == "__main__":
     begin_time = time.time()
 
-    print("base_dir: " + BASE_DIR)
-    print("models_dir: " + MODELS_DIR)
-
     main()
 
     print("total cost: " + str(time.time() - begin_time))

src/utils/images.py

 import io
 import json
-import os
 import time
 import traceback
 
 import cv2
-import dlib
 import numpy as np
 import requests
 from PIL import Image
 
-BASE_DIR = os.getcwd()
-MODELS_DIR = os.path.join(BASE_DIR, "_models")
-FACEREC_PATH = os.path.join(MODELS_DIR, "dlib_face_recognition_resnet_model_v1.dat")
-SHAPE_PATH = os.path.join(MODELS_DIR, "shape_predictor_68_face_landmarks.dat")
-
-dlib.DLIB_USE_CUDA = False
-
-face_rec = dlib.face_recognition_model_v1(FACEREC_PATH)
-face_detector = dlib.get_frontal_face_detector()
-shape_predictor = dlib.shape_predictor(SHAPE_PATH)
-
 
 def url_to_ndarray(url):
     result = requests.get(url, timeout=30)
@@ -41,7 +28,7 @@ def file_to_ndarray(path):
     return data
 
 
-def face_to_vec(img, max_size=700):
+def face_to_vec(img, face_rec_f, face_detector_f, shape_predictor_f, max_size=700):
     start = time.time()
     orig_img = img
     print("detect image...")
@@ -54,7 +41,7 @@ def face_to_vec(img, max_size=700):
         img = cv2.resize(img, size, interpolation=cv2.INTER_AREA)
 
     try:
-        dets = face_detector(img, 1)
+        dets = face_detector_f(img, 1)
     except Exception as e:
         print(e)
         traceback.print_exc()
@@ -67,14 +54,14 @@ def face_to_vec(img, max_size=700):
     for i, d in enumerate(dets):
         face = {}
 
-        shape = shape_predictor(img, d)
+        shape = shape_predictor_f(img, d)
         rect = shape.rect
         landmark = extract_landmark(shape, scale=scale)
         face["rect"] = check_rect(rect, orig_img.shape, scale=scale, landmark=landmark)
         face["landmark"] = json.dumps(landmark)
 
         try:
-            face_descriptor = face_rec.compute_face_descriptor(img, shape)
+            face_descriptor = face_rec_f.compute_face_descriptor(img, shape)
             face["feature"] = json.dumps(np.array(face_descriptor).tolist())
             faces.append(face)
             # del face_descriptor