app.py

import os
import cv2
import subprocess

subprocess.run(["sed", "-i", "9s/.*/import torchvision.transforms.functional as F_t/", "/usr/local/lib/python3.11/dist-packages/pytorchvideo/transforms/augmentations.py"], check=True)

import torch
from basicsr.archs.rrdbnet_arch import RRDBNet
from basicsr.utils.download_util import load_file_from_url
from realesrgan import RealESRGANer
from realesrgan.archs.srvgg_arch import SRVGGNetCompact
import gradio as gr

# 模型配置
MODEL_OPTIONS = {
    "RealESRGAN_x4plus": {
        "model": lambda: RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=4),
        "netscale": 4,
        "file_url": "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth"
    },
    "RealESRNet_x4plus": {
        "model": lambda: RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=4),
        "netscale": 4,
        "file_url": "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.1/RealESRNet_x4plus.pth"
    },
    "RealESRGAN_x4plus_anime_6B": {
        "model": lambda: RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=6, num_grow_ch=32, scale=4),
        "netscale": 4,
        "file_url": "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth"
    },
    "RealESRGAN_x2plus": {
        "model": lambda: RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=2),
        "netscale": 2,
        "file_url": "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.1/RealESRGAN_x2plus.pth"
    },
    "realesr-animevideov3": {
        "model": lambda: SRVGGNetCompact(num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=16, upscale=4, act_type='prelu'),
        "netscale": 4,
        "file_url": "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-animevideov3.pth"
    },
    "realesr-general-x4v3": {
        "model": lambda: SRVGGNetCompact(num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=32, upscale=4, act_type='prelu'),
        "netscale": 4,
        "file_url": "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-x4v3.pth"
    }
}

def load_model(model_name):
    """加载并初始化 Real-ESRGAN 模型"""
    model_config = MODEL_OPTIONS[model_name]
    model = model_config["model"]()
    netscale = model_config["netscale"]
    file_url = model_config["file_url"]

    # 下载模型权重
    model_path = os.path.join("weights", f"{model_name}.pth")
    if not os.path.isfile(model_path):
        os.makedirs("weights", exist_ok=True)
        model_path = load_file_from_url(url=file_url, model_dir="weights", progress=True, file_name=None)

    # 初始化 RealESRGANer
    upsampler = RealESRGANer(
        scale=netscale,
        model_path=model_path,
        model=model,
        tile=0,  # 默认无分块
        tile_pad=10,
        pre_pad=0,
        half=True  # 默认使用 fp16
    )
    return upsampler

def enhance_image(input_image, model_name, outscale, face_enhance):
    """执行图像超分辨率增强"""
    # 将 Gradio 上传的图像转换为 OpenCV 格式
    img = cv2.cvtColor(input_image, cv2.COLOR_RGB2BGR)

    # 确定图像模式
    if len(img.shape) == 3 and img.shape[2] == 4:
        img_mode = 'RGBA'
    else:
        img_mode = None

    # 加载模型
    upsampler = load_model(model_name)

    # 是否使用人脸增强
    if face_enhance:
        from gfpgan import GFPGANer
        face_enhancer = GFPGANer(
            model_path='https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.3.pth',
            upscale=outscale,
            arch='clean',
            channel_multiplier=2,
            bg_upsampler=upsampler
        )
        _, _, output = face_enhancer.enhance(img, has_aligned=False, only_center_face=False, paste_back=True)
    else:
        output, _ = upsampler.enhance(img, outscale=outscale)

    # 将结果转换回 RGB 格式以供 Gradio 显示
    output = cv2.cvtColor(output, cv2.COLOR_BGR2RGB)
    return output

# Gradio 界面
with gr.Blocks(title="Real-ESRGAN 图像超分辨率") as app:
    gr.Markdown("## Real-ESRGAN 图像超分辨率系统")
    gr.Markdown("上传图像，选择模型和参数，生成高清图像！")

    with gr.Row():
        with gr.Column():
            input_image = gr.Image(label="输入图像", type="numpy")
            model_dropdown = gr.Dropdown(
                choices=list(MODEL_OPTIONS.keys()),
                label="选择模型",
                value="RealESRGAN_x4plus"
            )
            outscale = gr.Slider(minimum=1, maximum=4, step=0.5, value=4, label="放大倍数")
            face_enhance = gr.Checkbox(label="启用人脸增强", value=False)
            enhance_btn = gr.Button("开始增强", variant="primary")

        with gr.Column():
            output_image = gr.Image(label="增强结果", type="numpy")

    enhance_btn.click(
        fn=enhance_image,
        inputs=[input_image, model_dropdown, outscale, face_enhance],
        outputs=[output_image],
        api_name="enhance"
    )

if __name__ == "__main__":
    app.launch(server_name="0.0.0.0", server_port=7860)