Upload 5 files

Dockerfile

@@ -0,0 +1,48 @@
+FROM nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04
+
+ARG DEBIAN_FRONTEND=noninteractive
+
+ENV PYTHONUNBUFFERED=1 \
+  ORT_CUDA_PROVIDER_OPTIONS=1
+
+#RUN apt-get install libcudnn8=${8.9.0}-1+${11.8.0} && apt-get install libcudnn8-dev=${8.9.0}-1+${11.8.0}
+
+RUN apt-get update && apt-get install --no-install-recommends -y \
+  build-essential \
+  python3.10-dev \
+  python3-pip \
+  git \
+  ffmpeg \
+  && apt-get clean && rm -rf /var/lib/apt/lists/*
+
+# Install NVIDIA CUDA Toolkit
+RUN apt-get update --yes && apt install nvidia-cuda-toolkit --yes && apt-get install zlib1g
+
+WORKDIR /code
+
+COPY ./requirements.txt /code/requirements.txt
+
+# Set up a new user named "user" with user ID 1000
+RUN useradd -m -u 1000 user
+# Switch to the "user" user
+USER user
+# Set home to the user's home directory
+ENV HOME=/home/user \
+	PATH=/home/user/.local/bin:$PATH \
+    PYTHONPATH=$HOME/app \
+	PYTHONUNBUFFERED=1 \
+	GRADIO_ALLOW_FLAGGING=never \
+	GRADIO_NUM_PORTS=1 \
+	GRADIO_SERVER_NAME=0.0.0.0 \
+	GRADIO_THEME=huggingface \
+	SYSTEM=spaces
+
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+
+# Set the working directory to the user's home directory
+WORKDIR $HOME/app
+
+# Copy the current directory contents into the container at $HOME/app setting the owner to the user
+COPY --chown=user . $HOME/app
+
+CMD ["python3", "app.py"]

GFPGANv1.4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2cd4703ab14f4d01fd1383a8a8b266f9a5833dacee8e6a79d3bf21a1b6be5ad
+size 348632874

app.py

@@ -0,0 +1,93 @@
+import cv2
+import os
+import insightface
+import onnxruntime
+from tqdm import tqdm
+import shutil
+import gfpgan
+import gradio as gr
+import subprocess
+from PIL import Image
+
+
+def video_to_frames(video_path, output_folder):
+    vidcap = cv2.VideoCapture(video_path)
+    fps = vidcap.get(cv2.CAP_PROP_FPS)
+    success, image = vidcap.read()
+    count = 0
+    if not os.path.exists(output_folder):
+        os.makedirs(output_folder)
+
+    while success:
+        frame_name = os.path.join(output_folder, f"frame_{count}.jpg")
+        cv2.imwrite(frame_name, image)
+        success, image = vidcap.read()
+        count += 1
+        if count>500:
+          break
+    print(f"{count} frames extracted from {video_path}.")
+    return [count,fps]
+
+def frames_to_video(frame_folder, video_path, image_path, frame_count,fps):
+    print("ImagePath",image_path)
+    frames = [f for f in os.listdir(frame_folder) if f.endswith('.jpg')]
+    frames.sort(key=lambda x: int(x.split('_')[1].split('.')[0]))  # Sort frames in ascending order
+
+    frame = cv2.imread(os.path.join(frame_folder, frames[0]))
+    height, width, _ = frame.shape
+
+    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
+    out = cv2.VideoWriter(video_path, fourcc, fps, (width, height))
+    
+    providers = ["CUDAExecutionProvider"]
+    print("Available Providers==",providers)
+    app = insightface.app.FaceAnalysis(name='buffalo_l', providers=providers)
+    app.prepare(ctx_id=0, det_size=(640, 640))
+    swapper = insightface.model_zoo.get_model("inswapper_128.onnx",download=False, download_zip=False,providers=providers)
+    face_enhancer = gfpgan.GFPGANer(model_path="GFPGANv1.4.pth", upscale=1, device='cuda')
+    
+    for i in tqdm(range(frame_count), desc="Converting frames to video"):
+        img1 = cv2.imread(os.path.join(frame_folder, frames[i]))
+        #img2_pil = Image.open(image_path)
+        #img2_cv2 = cv2.cvtColor(np.array(img2_pil), cv2.COLOR_RGB2BGR)
+
+        faces1 = app.get(img1)
+        for _ in range(20):
+            faces2 = app.get(image_path)
+            if faces2:
+                break
+        else:
+            return
+        if faces1:
+            face1 = faces1[0]
+            face2 = faces2[0]
+            result = img1.copy()
+            result = swapper.get(result, face1, face2, paste_back=True)
+            _, _, result = face_enhancer.enhance(result)
+            out.write(result)
+        else:
+            out.write(img1)
+        progress = int((i + 1) / frame_count * 100)
+        print(progress)
+    out.release()
+    
+    print(f"Video saved at {video_path}.")
+
+def face_swap(video_path, image_path):
+    output_folder = "Out_Frames"
+    frame_count = video_to_frames(video_path, output_folder)
+    if frame_count[0] > 150:
+        frame_count[0] = 150
+    output_video_path = "output_video.mp4" 
+    frames_to_video(output_folder, output_video_path, image_path, frame_count[0],frame_count[1])
+    return output_video_path 
+    
+
+iface = gr.Interface(
+    fn=face_swap,
+    inputs=["video", "image"],
+    outputs="video",
+    title="Face Swap",
+    description="Upload a video and an image. The faces in the video will be swapped with the face in the image.",
+)
+iface.launch(share=True)

inswapper_128.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4a3f08c753cb72d04e10aa0f7dbe3deebbf39567d4ead6dce08e98aa49e16af
+size 554253681

requirements.txt

@@ -0,0 +1,9 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+insightface
+onnxruntime-gpu==1.17.1
+opennsfw2
+keras
+gfpgan
+accelerate
+torchvision==0.16.2
+gradio