app.py

import argparse
import os
import yaml
import torch
import torch.backends.cudnn as cudnn
import numpy as np
from data_web import web_input
from models import DenoisingDiffusion, DiffusiveRestoration
import utils
import PIL.Image as Image
from torchvision.utils import make_grid
import gradio as gr
import tempfile

# tempfile.tempdir = "/home/dachuang/gradio/tmp/"

os.environ["CUDA_VISIBLE_DEVICES"] = "-1"

title_markdown = ("""
欢迎来到甲骨文文字演变模拟器
你只需要输入一张甲骨文图片，就可以看到它在不同随机种子下演变到汉字的结果。
输入的甲骨文保证是一个完整的文字和背景，不要有其他干扰。
""")


def config_get():
    parser = argparse.ArgumentParser()
    # 参数配置文件路径
    parser.add_argument("--config", default='config_web.yml', type=str, required=False, help="Path to the config file")
    args = parser.parse_args()

    with open(os.path.join(args.config), "r") as f:
        config = yaml.safe_load(f)
    new_config = dict2namespace(config)

    return new_config


def dict2namespace(config):
    namespace = argparse.Namespace()
    for key, value in config.items():
        if isinstance(value, dict):
            new_value = dict2namespace(value)
        else:
            new_value = value
        setattr(namespace, key, new_value)
    return namespace


config = config_get()
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
print("=> using device: {}".format(device))
config.device = device
# torch.backends.cudnn.benchmark = True

diffusion = DenoisingDiffusion(config, test=True)
model = DiffusiveRestoration(diffusion, config)
print("=> creating diffusion model")


# 加载数据
# seed = 61
# torch.manual_seed(seed)
# np.random.seed(seed)
# if torch.cuda.is_available():
#     torch.cuda.manual_seed_all(seed)
# image_path = "/home/dachuang/hsguan/JGWNET/壴30.png"
# img = Image.open(image_path).convert('RGB')
# image_web = np.array(img)
# image = web_input(image_web)

# output_image = model.web_restore(image, r=config.data.grid_r)
# utils.logging.save_image(output_image, 'web/0.png')
# grid = make_grid(output_image)
# ndarr = grid.mul(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).to("cpu", torch.uint8).numpy()


# im = Image.fromarray(ndarr)
# im.save('web_output.png', format=None)

def sepia(image_web, seed):
    torch.manual_seed(seed)
    np.random.seed(seed)
    if torch.cuda.is_available():
        torch.cuda.manual_seed_all(seed)
    image, image_tmp = web_input(image_web)
    print("开始生成")
    output_image = model.web_restore(image, r=config.data.grid_r)
    grid = make_grid(output_image)
    ndarr = grid.mul(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).to("cpu", torch.uint8).numpy()
    # torch.cuda.empty_cache()
    print("生成完成")
    return ndarr, image_tmp


demo = gr.Interface(sepia,
                    inputs=[gr.Image(label="输入甲骨文图片", height=600, width=600),
                            gr.Number(label="随机种子", value=61)],
                    outputs=[gr.Image(label="输出汉字图片", height=600, width=600),
                             gr.Image(label="矫正后甲骨文图片", height=600, width=600)],
                    title=title_markdown)
demo.launch(
    # server_name="127.0.0.1",
    # server_port=7681,
    share=True
)