glam-app

.gitignore

@@ -0,0 +1 @@
+/gg

Dockerfile

@@ -0,0 +1,15 @@
+FROM python:3.9
+
+WORKDIR /app
+
+# Copy the current directory contents into the container at /app
+COPY . /app
+
+# Install any needed packages specified in requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Make port 80 available to the world outside this container
+EXPOSE 7860
+
+# Run app.py when the container launches
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,8 +1,8 @@
 ---
-title: Glam Ai
-emoji: 😻
-colorFrom: green
-colorTo: purple
+title: Media Enhancer
+emoji: 🐢
+colorFrom: yellow
+colorTo: blue
 sdk: docker
 pinned: false
 license: mit

app.py

@@ -0,0 +1,74 @@
+from fastapi import FastAPI, File, UploadFile, HTTPException
+from fastapi.responses import StreamingResponse
+from image_enhancer import EnhancementMethod, Enhancer
+from video_enhancer import VideoEnhancer
+from pydantic import BaseModel
+from PIL import Image
+from io import BytesIO
+import base64
+import magic
+from typing import List
+class EnhancementRequest(BaseModel):
+    method: EnhancementMethod = EnhancementMethod.gfpgan
+    background_enhancement: bool = True
+    upscale: int = 2
+
+class _EnhanceBase(BaseModel):
+    encoded_base_img: List[str]
+
+
+
+app = FastAPI()
+
+@app.get("/")
+def greet_json():
+    return {"Initializing GlamApp Enhancer"}
+
+@app.post("/enhance/image/")
+async def enhance_image(
+    file: UploadFile = File(...),
+    request: EnhancementRequest = EnhancementRequest()
+):
+    try:
+        if not file.content_type.startswith('image/'):
+            raise HTTPException(status_code=400, detail="Invalid file type")
+            
+        contents = await file.read()
+        base64_encoded_image = base64.b64encode(contents).decode('utf-8')
+        data = _EnhanceBase(encoded_base_img=[base64_encoded_image])
+
+        enhancer = Enhancer(request.method, request.background_enhancement, request.upscale)
+
+        enhanced_img, original_resolution, enhanced_resolution = await enhancer.enhance(data)
+
+        enhanced_image = Image.fromarray(enhanced_img)
+        img_byte_arr = BytesIO()
+        enhanced_image.save(img_byte_arr, format='PNG')
+        img_byte_arr.seek(0)
+        print(original_resolution, enhanced_resolution)
+        
+        return StreamingResponse(img_byte_arr, media_type="image/png")
+
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+        
+@app.post("/enhance/video/")
+async def enhance_video(file: UploadFile = File(...)):
+    enhancer = VideoEnhancer()
+    file_header = await file.read(1024)
+    file.file.seek(0)
+    mime = magic.Magic(mime=True)
+    file_mime_type = mime.from_buffer(file_header)
+    
+    accepted_mime_types = [
+        'video/mp4',
+        'video/mpeg',
+        'video/x-msvideo',
+        'video/quicktime',
+        'video/x-matroska',
+        'video/webm'
+    ]
+    
+    if file_mime_type not in accepted_mime_types:
+        raise HTTPException(status_code=400, detail="Invalid file type. Please upload a video file.")
+    return await enhancer.stream_enhanced_video(file)

image_enhancer.py

@@ -0,0 +1,123 @@
+import os
+import torch 
+from gfpgan import GFPGANer
+from tqdm import tqdm
+import cv2
+from realesrgan import RealESRGANer
+from basicsr.archs.rrdbnet_arch import RRDBNet
+import warnings
+from enum import Enum
+
+class EnhancementMethod(str, Enum):
+    gfpgan = "gfpgan"
+    RestoreFormer = "RestoreFormer"
+    codeformer = "codeformer"
+    realesrgan = "realesrgan"
+
+
+class Enhancer:
+    def __init__(self, method: EnhancementMethod, background_enhancement=True, upscale=2):
+        self.method = method
+        self.background_enhancement = background_enhancement
+        self.upscale = upscale
+        self.bg_upsampler = None
+        self.realesrgan_enhancer = None
+
+        if self.method != EnhancementMethod.realesrgan:
+            self.setup_face_enhancer()
+            if self.background_enhancement:
+                self.setup_background_enhancer()
+        else:
+            self.setup_realesrgan_enhancer()
+
+    def setup_background_enhancer(self):
+        if not torch.cuda.is_available():
+            warnings.warn('The unoptimized RealESRGAN is slow on CPU. We do not use it.')
+            return
+
+        model = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=self.upscale)
+        model_path = f'https://huggingface.co/dtarnow/UPscaler/resolve/main/RealESRGAN_x{self.upscale}plus.pth'
+        self.bg_upsampler = RealESRGANer(
+            scale=self.upscale,
+            model_path=model_path,
+            model=model,
+            tile=400,
+            tile_pad=10,
+            pre_pad=0,
+            half=True)
+
+    def setup_realesrgan_enhancer(self):
+        if not torch.cuda.is_available():
+            raise ValueError('CUDA is not available for RealESRGAN')
+
+        model = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=self.upscale)
+        model_path = f'https://huggingface.co/dtarnow/UPscaler/resolve/main/RealESRGAN_x{self.upscale}plus.pth'
+        self.realesrgan_enhancer = RealESRGANer(
+            scale=self.upscale,
+            model_path=model_path,
+            model=model,
+            tile=400,
+            tile_pad=10,
+            pre_pad=0,
+            half=True)
+
+    def setup_face_enhancer(self):
+        model_configs = {
+            EnhancementMethod.gfpgan: {
+                'arch': 'clean',
+                'channel_multiplier': 2,
+                'model_name': 'GFPGANv1.4',
+                'url': 'https://huggingface.co/gmk123/GFPGAN/resolve/main/GFPGANv1.4.pth'
+            },
+            EnhancementMethod.RestoreFormer: {
+                'arch': 'RestoreFormer',
+                'channel_multiplier': 2,
+                'model_name': 'RestoreFormer',
+                'url': 'https://github.com/TencentARC/GFPGAN/releases/download/v1.3.4/RestoreFormer.pth'
+            },
+            EnhancementMethod.codeformer: {
+                'arch': 'CodeFormer',
+                'channel_multiplier': 2,
+                'model_name': 'CodeFormer',
+                'url': 'https://huggingface.co/sinadi/aar/resolve/main/codeformer.pth'
+            }
+        }
+
+        config = model_configs.get(self.method)
+        if not config:
+            raise ValueError(f'Wrong model version {self.method}')
+
+        model_path = os.path.join('gfpgan/weights', config['model_name'] + '.pth')
+        if not os.path.isfile(model_path):
+            model_path = os.path.join('checkpoints', config['model_name'] + '.pth')
+        if not os.path.isfile(model_path):
+            model_path = config['url']
+
+        self.face_enhancer = GFPGANer(
+            model_path=model_path,
+            upscale=self.upscale,
+            arch=config['arch'],
+            channel_multiplier=config['channel_multiplier'],
+            bg_upsampler=self.bg_upsampler)
+
+    def check_image_resolution(self, image):
+        height, width, _ = image.shape
+        return width, height
+
+    async def enhance(self, image):
+        img = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+        width, height = self.check_image_resolution(img)
+        
+        if self.method == EnhancementMethod.realesrgan:
+            enhanced_img, _ = await asyncio.to_thread(self.realesrgan_enhancer.enhance, img, outscale=self.upscale)
+        else:
+            _, _, enhanced_img = await asyncio.to_thread(self.face_enhancer.enhance,
+                img,
+                has_aligned=False,
+                only_center_face=False,
+                paste_back=True)
+        
+        enhanced_img = cv2.cvtColor(enhanced_img, cv2.COLOR_BGR2RGB)
+        enhanced_width, enhanced_height = self.check_image_resolution(enhanced_img)
+
+        return enhanced_img, (width, height), (enhanced_width, enhanced_height)

requirements.txt

@@ -0,0 +1,13 @@
+fastapi
+uvicorn[standard]
+gfpgan==1.3.8
+realesrgan==0.3.0
+pillow==10.3.0
+pydantic==2.7.1
+pydantic-settings==2.0.3
+pydantic_core==2.18.2
+requests==2.31.0
+basicsr
+huggingface-hub==0.25.1
+numpy==1.26.4
+facexlib==0.3.0

video_enhancer.py

@@ -0,0 +1,69 @@
+import cv2
+import numpy as np
+import torch
+import io
+import asyncio
+from fastapi.responses import StreamingResponse
+from basicsr.archs.rrdbnet_arch import RRDBNet
+from realesrgan import RealESRGANer
+from huggingface_hub import hf_hub_download
+from concurrent.futures import ThreadPoolExecutor
+
+class VideoEnhancer:
+    def __init__(self, model_name="RealESRGAN_x4plus"):
+        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+        self.model = self.load_model(model_name)
+        self.executor = ThreadPoolExecutor(max_workers=4)
+
+    def load_model(self, model_name):
+        if model_name == "RealESRGAN_x4plus":
+            model = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=4)
+            model_path = hf_hub_download("schwgHao/RealESRGAN_x4plus", "RealESRGAN_x4plus.pth")
+            return RealESRGANer(scale=4, model_path=model_path, model=model, tile=0, tile_pad=10, pre_pad=0, half=True)
+        else:
+            raise ValueError(f"Unsupported model: {model_name}")
+
+    async def enhance_frame(self, frame):
+        loop = asyncio.get_running_loop()
+      
+        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+       
+        enhanced, _ = await loop.run_in_executor(self.executor, self.model.enhance, frame_rgb)
+        
+        return cv2.cvtColor(enhanced, cv2.COLOR_RGB2BGR)
+
+    async def process_video(self, input_bytes, output_bytes):
+        cap = cv2.VideoCapture(input_bytes)
+        width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+        height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+        fps = cap.get(cv2.CAP_PROP_FPS)
+
+        fourcc = cv2.VideoWriter_fourcc(*'mp4v')
+        out = cv2.VideoWriter(output_bytes, fourcc, fps, (width * 4, height * 4))
+
+        while cap.isOpened():
+            ret, frame = cap.read()
+            if not ret:
+                break
+            enhanced_frame = await self.enhance_frame(frame)
+            out.write(enhanced_frame)
+
+        cap.release()
+        out.release()
+
+    async def stream_enhanced_video(self, video_file):
+        video_bytes = await video_file.read()
+        cap = cv2.VideoCapture(io.BytesIO(video_bytes).getvalue())
+        
+        async def generate():
+            while cap.isOpened():
+                ret, frame = cap.read()
+                if not ret:
+                    break
+                enhanced_frame = await self.enhance_frame(frame)
+                _, buffer = cv2.imencode('.jpg', enhanced_frame)
+                yield (b'--frame\r\n'
+                       b'Content-Type: image/jpeg\r\n\r\n' + buffer.tobytes() + b'\r\n')
+            cap.release()
+
+        return StreamingResponse(generate(), media_type="multipart/x-mixed-replace; boundary=frame")