Upload 19 files

.gitattributes

@@ -33,3 +33,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+FINAL-EFFICIENTNETV2-B0/variables/variables.data-00000-of-00001 filter=lfs diff=lfs merge=lfs -text
+obama-fake.mp4 filter=lfs diff=lfs merge=lfs -text
+Video1-fake-1-ff.mp4 filter=lfs diff=lfs merge=lfs -text
+Video3-fake-3-ff.mp4 filter=lfs diff=lfs merge=lfs -text
+Video6-real-1-ff.mp4 filter=lfs diff=lfs merge=lfs -text
+Video8-real-3-ff.mp4 filter=lfs diff=lfs merge=lfs -text

FINAL-EFFICIENTNETV2-B0.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2006d2344a1804c269dd6c686181ee0e28213db5c52b73abfbfacf99e0d604e2
+size 22914217

FINAL-EFFICIENTNETV2-B0/keras_metadata.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8411f85bd22de246fee31adc6bbf0a60d403ac22d8f572154fd77eb866b8daf3
+size 202114

FINAL-EFFICIENTNETV2-B0/saved_model.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca17aff86eeedbeab2ace0fc42296a1fe11352c6adb418f04f96c5a3607bd28a
+size 10505251

FINAL-EFFICIENTNETV2-B0/variables/variables.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6418ccca9c8b62339ccfae9e5e3aae785fbdeed31fa08af7207ad4f0fc94fbbf
+size 23824720

Video1-fake-1-ff.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58262ed5e804069587e393ed06b48e655ca35d7ad58b68c161f5356a14482c48
+size 1746578

Video3-fake-3-ff.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e9ef4e65483c4152f477803fb083be53ec9311e4006abaacbf9647bf3ae0fa5
+size 9101725

Video6-real-1-ff.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad4e54db5f1b0c2f556e039d61ec38e7195edbba6257e266244be64af0bda5e3
+size 1771036

Video8-real-3-ff.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:719f49698458abfa2ff25eb617ff03c5e56ddea51d912d65fbfa44c3db94768a
+size 8949516

app.py

@@ -0,0 +1,179 @@
+import gradio as gr
+import cv2
+import numpy as np
+import tensorflow as tf
+import tensorflow_addons
+
+from facenet_pytorch import MTCNN
+from PIL import Image
+import moviepy.editor as mp
+import os
+import zipfile
+
+local_zip = "FINAL-EFFICIENTNETV2-B0.zip"
+zip_ref = zipfile.ZipFile(local_zip, 'r')
+zip_ref.extractall('FINAL-EFFICIENTNETV2-B0')
+zip_ref.close()
+
+# Load face detector
+mtcnn = MTCNN(margin=14, keep_all=True, factor=0.7, device='cpu')
+
+#Face Detection function, Reference: (Timesler, 2020); Source link: https://www.kaggle.com/timesler/facial-recognition-model-in-pytorch
+class DetectionPipeline:
+    """Pipeline class for detecting faces in the frames of a video file."""
+
+    def __init__(self, detector, n_frames=None, batch_size=60, resize=None):
+        """Constructor for DetectionPipeline class.
+
+        Keyword Arguments:
+            n_frames {int} -- Total number of frames to load. These will be evenly spaced
+                throughout the video. If not specified (i.e., None), all frames will be loaded.
+                (default: {None})
+            batch_size {int} -- Batch size to use with MTCNN face detector. (default: {32})
+            resize {float} -- Fraction by which to resize frames from original prior to face
+                detection. A value less than 1 results in downsampling and a value greater than
+                1 result in upsampling. (default: {None})
+        """
+        self.detector = detector
+        self.n_frames = n_frames
+        self.batch_size = batch_size
+        self.resize = resize
+
+    def __call__(self, filename):
+        """Load frames from an MP4 video and detect faces.
+
+        Arguments:
+            filename {str} -- Path to video.
+        """
+        # Create video reader and find length
+        v_cap = cv2.VideoCapture(filename)
+        v_len = int(v_cap.get(cv2.CAP_PROP_FRAME_COUNT))
+
+        # Pick 'n_frames' evenly spaced frames to sample
+        if self.n_frames is None:
+            sample = np.arange(0, v_len)
+        else:
+            sample = np.linspace(0, v_len - 1, self.n_frames).astype(int)
+
+        # Loop through frames
+        faces = []
+        frames = []
+        for j in range(v_len):
+            success = v_cap.grab()
+            if j in sample:
+                # Load frame
+                success, frame = v_cap.retrieve()
+                if not success:
+                    continue
+                frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+                # frame = Image.fromarray(frame)
+
+                # Resize frame to desired size
+                if self.resize is not None:
+                    frame = frame.resize([int(d * self.resize) for d in frame.size])
+                frames.append(frame)
+
+                # When batch is full, detect faces and reset frame list
+                if len(frames) % self.batch_size == 0 or j == sample[-1]:
+
+                    boxes, probs = self.detector.detect(frames)
+
+                    for i in range(len(frames)):
+
+                        if boxes[i] is None:
+                            faces.append(face2)     #append previous face frame if no face is detected
+                            continue
+
+                        box = boxes[i][0].astype(int)
+                        frame = frames[i]
+                        face = frame[box[1]:box[3], box[0]:box[2]]
+
+                        if not face.any():
+                            faces.append(face2)     #append previous face frame if no face is detected
+                            continue
+
+                        face2 = cv2.resize(face, (224, 224))
+
+                        faces.append(face2)
+
+                    frames = []
+
+        v_cap.release()
+
+        return faces
+
+
+detection_pipeline = DetectionPipeline(detector=mtcnn,n_frames=20, batch_size=60)
+
+model = tf.keras.models.load_model("FINAL-EFFICIENTNETV2-B0")
+
+
+def deepfakespredict(input_video):
+
+    faces = detection_pipeline(input_video)
+
+    total = 0
+    real = 0
+    fake = 0
+
+    for face in faces:
+
+        face2 = face/255
+        pred = model.predict(np.expand_dims(face2, axis=0))[0]
+        total+=1
+
+        pred2 = pred[1]
+
+        if pred2 > 0.5:
+          fake+=1
+        else:
+          real+=1
+
+    fake_ratio = fake/total
+
+    text =""
+    text2 = "Deepfakes Confidence: " + str(fake_ratio*100) + "%"
+
+    if fake_ratio >= 0.5:
+        text = "The video is FAKE."
+    else:
+        text = "The video is REAL."
+
+    face_frames = []
+    
+    for face in faces:
+        face_frame = Image.fromarray(face.astype('uint8'), 'RGB')
+        face_frames.append(face_frame)
+        
+    face_frames[0].save('results.gif', save_all=True, append_images=face_frames[1:], duration = 250, loop = 100 )
+    clip = mp.VideoFileClip("results.gif")
+    clip.write_videofile("video.mp4")
+
+    return text, text2, "video.mp4"
+
+
+
+title="EfficientNetV2 Deepfakes Video Detector"
+description="This is a demo implementation of EfficientNetV2 Deepfakes Image Detector by using frame-by-frame detection. \
+            To use it, simply upload your video, or click one of the examples to load them.\
+            This demo and model represent the Final Year Project titled \"Achieving Face Swapped Deepfakes Detection Using EfficientNetV2\" by a CS undergraduate Lee Sheng Yeh. \
+            The examples were extracted from Celeb-DF(V2)(Li et al, 2020) and FaceForensics++(Rossler et al., 2019). Full reference details is available in \"references.txt.\" \
+            The examples are used under fair use to demo the working of the model only. If any copyright is infringed, please contact the researcher via this email: tp054565@mail.apu.edu.my.\
+            "
+            
+examples = [              
+                ['Video1-fake-1-ff.mp4'],
+                ['Video6-real-1-ff.mp4'],
+                ['Video3-fake-3-ff.mp4'],
+                ['Video8-real-3-ff.mp4'],
+                ['real-1.mp4'],
+                ['fake-1.mp4'],
+           ]
+           
+gr.Interface(deepfakespredict,
+                     inputs = ["video"],
+                     outputs=["text","text", gr.outputs.Video(label="Detected face sequence")],
+                     title=title,
+                     description=description,
+                     examples=examples
+                     ).launch()

obama-fake.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c3812d4b9baabc09d68a7a02601ae69c367ac49826b9a2e5cbdcbca59b843f1
+size 4610918

packages.txt

@@ -0,0 +1,3 @@
+ffmpeg 
+libsm6 
+libxext6

references.txt

@@ -0,0 +1,27 @@
+Dataset References
+
+Celeb-DF (V2)
+
+@inproceedings{Celeb_DF_cvpr20, 
+author = {Yuezun Li and Xin Yang and Pu Sun and Honggang Qi and Siwei Lyu}, 
+title = {Celeb-DF: A Large-scale Challenging Dataset for DeepFake Forensics}, 
+booktitle= {IEEE Conference on Computer Vision and Patten Recognition (CVPR)}, 
+year = {2020}} 
+
+
+FaceForensics++ Dataset 
+
+@inproceedings{roessler2019faceforensicspp, 
+author = {Andreas R\"ossler and Davide Cozzolino and Luisa Verdoliva and Christian Riess and Justus Thies and Matthias Nie{\ss}ner}, 
+title = {Face{F}orensics++: Learning to Detect Manipulated Facial Images}, 
+booktitle= {International Conference on Computer Vision (ICCV)}, 
+year = {2019} }
+
+@MISC{DDD_GoogleJigSaw2019, 
+AUTHOR = {Dufour, Nicholas and Gully, Andrew and Karlsson, Per and Vorbyov, Alexey Victor and Leung, Thomas and Childs, Jeremiah and Bregler, Christoph}, 
+DATE = {2019-09}, 
+TITLE = {DeepFakes Detection Dataset by Google & JigSaw}}
+
+
+Face Detection Fucntion (Timesler, 2020)
+Source Link: https://www.kaggle.com/timesler/facial-recognition-model-in-pytorch

requirements.txt

@@ -0,0 +1,8 @@
+tensorflow
+tensorflow-addons
+facenet_pytorch
+numpy
+opencv-python
+opencv-python-headless
+mtcnn
+moviepy