Update refacer.py

refacer.py

@@ -24,20 +24,20 @@ import re
 import subprocess
 
 class RefacerMode(Enum):
-     CPU, CUDA, COREML, TENSORRT = range(1, 5)
+    CPU, CUDA, COREML, TENSORRT = range(1, 5)
 
 class Refacer:
-    def __init__(self,force_cpu=False,colab_performance=False):
+    def __init__(self, force_cpu=False, colab_performance=False):
         self.first_face = False
         self.force_cpu = force_cpu
         self.colab_performance = colab_performance
-        self.__check_encoders()
+        self.__check_encoders()  # Correct method name
         self.__check_providers()
         self.total_mem = psutil.virtual_memory().total
         self.__init_apps()
 
     def __check_providers(self):
-        if self.force_cpu :
+        if self.force_cpu:
             self.providers = ['CPUExecutionProvider']
         else:
             self.providers = rt.get_available_providers()
@@ -48,18 +48,18 @@ class Refacer:
 
         if len(self.providers) == 1 and 'CPUExecutionProvider' in self.providers:
             self.mode = RefacerMode.CPU
-            self.use_num_cpus = mp.cpu_count()-1
-            self.sess_options.intra_op_num_threads = int(self.use_num_cpus/3)
+            self.use_num_cpus = mp.cpu_count() - 1
+            self.sess_options.intra_op_num_threads = int(self.use_num_cpus / 3)
             print(f"CPU mode with providers {self.providers}")
         elif self.colab_performance:
             self.mode = RefacerMode.TENSORRT
-            self.use_num_cpus = mp.cpu_count()-1
-            self.sess_options.intra_op_num_threads = int(self.use_num_cpus/3)
+            self.use_num_cpus = mp.cpu_count() - 1
+            self.sess_options.intra_op_num_threads = int(self.use_num_cpus / 3)
             print(f"TENSORRT mode with providers {self.providers}")
         elif 'CoreMLExecutionProvider' in self.providers:
             self.mode = RefacerMode.COREML
-            self.use_num_cpus = mp.cpu_count()-1
-            self.sess_options.intra_op_num_threads = int(self.use_num_cpus/3)
+            self.use_num_cpus = mp.cpu_count() - 1
+            self.sess_options.intra_op_num_threads = int(self.use_num_cpus / 3)
             print(f"CoreML mode with providers {self.providers}")
         elif 'CUDAExecutionProvider' in self.providers:
             self.mode = RefacerMode.CUDA
@@ -74,25 +74,25 @@ class Refacer:
 
         model_path = os.path.join(assets_dir, 'det_10g.onnx')
         sess_face = rt.InferenceSession(model_path, self.sess_options, providers=self.providers)
-        self.face_detector = SCRFD(model_path,sess_face)
-        self.face_detector.prepare(0,input_size=(640, 640))
+        self.face_detector = SCRFD(model_path, sess_face)
+        self.face_detector.prepare(0, input_size=(640, 640))
 
-        model_path = os.path.join(assets_dir , 'w600k_r50.onnx')
+        model_path = os.path.join(assets_dir, 'w600k_r50.onnx')
         sess_rec = rt.InferenceSession(model_path, self.sess_options, providers=self.providers)
-        self.rec_app = ArcFaceONNX(model_path,sess_rec)
+        self.rec_app = ArcFaceONNX(model_path, sess_rec)
         self.rec_app.prepare(0)
 
         model_path = 'inswapper_128.onnx'
         sess_swap = rt.InferenceSession(model_path, self.sess_options, providers=self.providers)
-        self.face_swapper = INSwapper(model_path,sess_swap)
+        self.face_swapper = INSwapper(model_path, sess_swap)
 
     def prepare_faces(self, faces):
-        self.replacement_faces=[]
+        self.replacement_faces = []
         for face in faces:
             if "origin" in face:
                 face_threshold = face['threshold']
-                bboxes1, kpss1 = self.face_detector.autodetect(face['origin'], max_num=1)  
-                if len(kpss1)<1:
+                bboxes1, kpss1 = self.face_detector.autodetect(face['origin'], max_num=1)
+                if len(kpss1) < 1:
                     raise Exception('No face detected on "Face to replace" image')
                 feat_original = self.rec_app.get(face['origin'], kpss1[0])
             else:
@@ -100,13 +100,13 @@ class Refacer:
                 self.first_face = True
                 feat_original = None
                 print('No origin image: First face change')
-            _faces = self.__get_faces(face['destination'],max_num=1)
-            if len(_faces)<1:
+            _faces = self.__get_faces(face['destination'], max_num=1)
+            if len(_faces) < 1:
                 raise Exception('No face detected on "Destination face" image')
-            self.replacement_faces.append((feat_original,_faces[0],face_threshold))
+            self.replacement_faces.append((feat_original, _faces[0], face_threshold))
 
-    def __get_faces(self,frame,max_num=0):
-        bboxes, kpss = self.face_detector.detect(frame,max_num=max_num,metric='default')
+    def __get_faces(self, frame, max_num=0):
+        bboxes, kpss = self.face_detector.detect(frame, max_num=max_num, metric='default')
 
         if bboxes.shape[0] == 0:
             return []
@@ -122,24 +122,24 @@ class Refacer:
             ret.append(face)
         return ret
 
-    def process_first_face(self,frame):
-        faces = self.__get_faces(frame,max_num=1)
+    def process_first_face(self, frame):
+        faces = self.__get_faces(frame, max_num=1)
         if len(faces) != 0:
             frame = self.face_swapper.get(frame, faces[0], self.replacement_faces[0][1], paste_back=True)
         return frame
 
-    def process_faces(self,frame):
-        faces = self.__get_faces(frame,max_num=0)
+    def process_faces(self, frame):
+        faces = self.__get_faces(frame, max_num=0)
         for rep_face in self.replacement_faces:
             for i in range(len(faces) - 1, -1, -1):
                 sim = self.rec_app.compute_sim(rep_face[0], faces[i].embedding)
-                if sim>=rep_face[2]:
+                if sim >= rep_face[2]:
                     frame = self.face_swapper.get(frame, faces[i], rep_face[1], paste_back=True)
                     del faces[i]
                     break
         return frame
 
-    def __check_video_has_audio(self,video_path):
+    def __check_video_has_audio(self, video_path):
         self.video_has_audio = False
         probe = ffmpeg.probe(video_path)
         audio_stream = next((stream for stream in probe['streams'] if stream['codec_type'] == 'audio'), None)
@@ -148,11 +148,11 @@ class Refacer:
 
     def reface_group(self, faces, frames):
         results = []
-        with ThreadPoolExecutor(max_workers = self.use_num_cpus) as executor:
+        with ThreadPoolExecutor(max_workers=self.use_num_cpus) as executor:
             if self.first_face:
-                results = list(tqdm(executor.map(self.process_first_face, frames), total=len(frames),desc="Processing frames"))
+                results = list(tqdm(executor.map(self.process_first_face, frames), total=len(frames), desc="Processing frames"))
             else:
-                results = list(tqdm(executor.map(self.process_faces, frames), total=len(frames),desc="Processing frames"))
+                results = list(tqdm(executor.map(self.process_faces, frames), total=len(frames), desc="Processing frames"))
         return results
 
     def reface(self, video_path, faces):
@@ -162,13 +162,13 @@ class Refacer:
         cap = cv2.VideoCapture(video_path)
         total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
         print(f"Total frames: {total_frames}")
-        
+
         fps = cap.get(cv2.CAP_PROP_FPS)
         frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
         frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
 
         frames = []
-        with tqdm(total=total_frames,desc="Extracting frames") as pbar:
+        with tqdm(total=total_frames, desc="Extracting frames") as pbar:
             while cap.isOpened():
                 flag, frame = cap.read()
                 if flag and len(frame) > 0:
@@ -181,7 +181,7 @@ class Refacer:
             pbar.close()
 
         refaced_frames = self.reface_group(faces, frames)
-        
+
         video_buffer = io.BytesIO()
         out = cv2.VideoWriter('temp.mp4', cv2.VideoWriter_fourcc(*'mp4v'), fps, (frame_width, frame_height))