Merge pull request #1319 from Circuit8/represent-optimizations

Represent optimizations

.gitignore

@@ -17,3 +17,4 @@ benchmarks/results
 benchmarks/outputs
 benchmarks/dataset
 benchmarks/lfwe
+venv

deepface/modules/detection.py

@@ -1,15 +1,16 @@
 # built-in dependencies
-from typing import Any, Dict, List, Tuple, Union
+from typing import Any, Dict, List, Tuple, Union, Optional
 
 # 3rd part dependencies
+from heapq import nlargest
 import numpy as np
 import cv2
-from PIL import Image
 
 # project dependencies
 from deepface.modules import modeling
 from deepface.models.Detector import Detector, DetectedFace, FacialAreaRegion
 from deepface.commons import image_utils
+
 from deepface.commons.logger import Logger
 
 logger = Logger()
@@ -27,6 +28,7 @@ def extract_faces(
     color_face: str = "rgb",
     normalize_face: bool = True,
     anti_spoofing: bool = False,
+    max_faces: Optional[int] = None,
 ) -> List[Dict[str, Any]]:
     """
     Extract faces from a given image
@@ -97,6 +99,7 @@ def extract_faces(
             img=img,
             align=align,
             expand_percentage=expand_percentage,
+            max_faces=max_faces,
         )
 
     # in case of no face found
@@ -176,7 +179,9 @@ def extract_faces(
 
 
 def detect_faces(
-    detector_backend: str, img: np.ndarray, align: bool = True, expand_percentage: int = 0
+    detector_backend: str, img: np.ndarray,
+    align: bool = True, expand_percentage: int = 0,
+    max_faces: Optional[int] = None
 ) -> List[DetectedFace]:
     """
     Detect face(s) from a given image
@@ -202,7 +207,6 @@ def detect_faces(
         - confidence (float): The confidence score associated with the detected face.
     """
     height, width, _ = img.shape
-
     face_detector: Detector = modeling.build_model(
         task="face_detector", model_name=detector_backend
     )
@@ -233,8 +237,29 @@ def detect_faces(
     # find facial areas of given image
     facial_areas = face_detector.detect_faces(img)
 
-    results = []
-    for facial_area in facial_areas:
+    if max_faces is not None and max_faces < len(facial_areas):
+        facial_areas = nlargest(
+            max_faces,
+            facial_areas,
+            key=lambda facial_area: facial_area.w * facial_area.h
+        )
+
+    return [
+        expand_and_align_face(
+            facial_area=facial_area,
+            img=img,
+            align=align,
+            expand_percentage=expand_percentage,
+            width_border=width_border,
+            height_border=height_border
+        )
+        for facial_area in facial_areas
+    ]
+
+def expand_and_align_face(
+    facial_area: FacialAreaRegion, img: np.ndarray,
+    align: bool, expand_percentage: int, width_border: int,
+    height_border: int) -> DetectedFace:
     x = facial_area.x
     y = facial_area.y
     w = facial_area.w
@@ -256,7 +281,6 @@ def detect_faces(
 
     # extract detected face unaligned
     detected_face = img[int(y) : int(y + h), int(x) : int(x + w)]
-
     # align original image, then find projection of detected face area after alignment
     if align is True:  # and left_eye is not None and right_eye is not None:
         aligned_img, angle = align_img_wrt_eyes(img=img, left_eye=left_eye, right_eye=right_eye)
@@ -277,16 +301,13 @@ def detect_faces(
         if right_eye is not None:
             right_eye = (right_eye[0] - width_border, right_eye[1] - height_border)
 
-        result = DetectedFace(
+    return DetectedFace(
         img=detected_face,
         facial_area=FacialAreaRegion(
             x=x, y=y, h=h, w=w, confidence=confidence, left_eye=left_eye, right_eye=right_eye
         ),
         confidence=confidence,
     )
-        results.append(result)
-    return results
-
 
 def align_img_wrt_eyes(
     img: np.ndarray,
@@ -311,7 +332,16 @@ def align_img_wrt_eyes(
         return img, 0
 
     angle = float(np.degrees(np.arctan2(left_eye[1] - right_eye[1], left_eye[0] - right_eye[0])))
-    img = np.array(Image.fromarray(img).rotate(angle, resample=Image.BICUBIC))
+
+    (h, w) = img.shape[:2]
+    center = (w // 2, h // 2)
+    M = cv2.getRotationMatrix2D(center, angle, 1.0)
+    img = cv2.warpAffine(
+        img, M, (w, h),
+        flags=cv2.INTER_CUBIC, borderMode=cv2.BORDER_CONSTANT,
+        borderValue=(0,0,0)
+    )
+
     return img, angle
 
 

deepface/modules/representation.py

@@ -81,6 +81,7 @@ def represent(
             align=align,
             expand_percentage=expand_percentage,
             anti_spoofing=anti_spoofing,
+            max_faces=max_faces,
         )
     else:  # skip
         # Try load. If load error, will raise exception internal