Update app.py

app.py

@@ -1,52 +1,53 @@
+
+
 from moviepy.editor import *
-import face_recognition
 from path import Path
 from PIL import Image, ImageDraw
 import numpy as np 
 
 import gradio 
 import os
-os.makedirs('video', exist_ok=True)
+from insightface.app import FaceAnalysis
+from insightface.data import get_image as ins_get_image
 
+os.makedirs('video', exist_ok=True)
 
 def procss_video(video_str):
     source_frames = []
-    # video_name = 'ZW9WPPMgpWP4bjjs.mp4'
+
     clip = VideoFileClip(video_str)
     for item in clip.iter_frames():
         source_frames.append(item)
 
     audioclip = clip.audio
-    im2 = Image.open("mask_output.png")
-
-
-
 
 
+    app = FaceAnalysis(providers=['CUDAExecutionProvider'])
+    app.prepare(ctx_id=0, det_size=(640, 640))
 
+    im2 = Image.open('mask_output.png')
     dealed_frames = []
-    batch_size = 8
-    for i in range(0, len(source_frames), batch_size):
-        frames = source_frames[i:i+batch_size]
-        # print('processing framese from {} to {}'.format(i, i+batch_size))
-        batch_of_face_locations = face_recognition.batch_face_locations(frames, number_of_times_to_upsample=2, batch_size = batch_size)
 
-        # Now let's list all the faces we found in all 128 frames
-        for frame_number_in_batch, face_locations in enumerate(batch_of_face_locations):
+    for item in source_frames:
+        pil_image = Image.fromarray(item)
+        faces = app.get(item)
+        # rimg = app.draw_on(img, faces)
+        for face in faces:
+            # print(face)
+            face.bbox = face.bbox.astype(np.int)
+            top ,  right,bottom,  left = face.bbox
+            #find top right bottom left from face.bbox
+            left, bottom, right, top = face.bbox
 
-            pil_image = Image.fromarray(frames[frame_number_in_batch])
-            draw = ImageDraw.Draw(pil_image)
-            # Loop through each face found in the unknown image
-            for (top, right, bottom, left) in face_locations:
+            im = im2.resize((int(abs(top-bottom)*0.8), int(abs(left-right)*0.8) ))
+            pil_image.paste(im, (left, int((top+bottom)/2)),  im)
 
-                im = im2.resize((int(abs(top-bottom)*0.8), int(abs(left-right)*0.8) ))
-                pil_image.paste(im, (left, int((top+bottom)/2)),  im)
-            dealed_frames.append(np.array(pil_image))
+        dealed_frames.append(np.array(pil_image))
 
-    output_clip = ImageSequenceClip(dealed_frames, fps=20)
+    output_clip = ImageSequenceClip(dealed_frames, fps = clip.fps)
     new_audioclip = CompositeAudioClip([audioclip])
     output_clip.audio = new_audioclip
-    output_clip.write_videofile(os.path.join('video', 'processed_'+Path(video_str).name), fps=20,codec="libx264", audio_codec="aac")
+    output_clip.write_videofile(os.path.join('video', 'processed_'+Path(video_str).name),codec="libx264", audio_codec="aac")
 
     return os.path.join('video', 'processed_'+Path(video_str).name)
 # os.mkdir('video')