Video_cutter/utils/rmbg.py

import os 
from PIL import Image
import torch
from torchvision import transforms
from transformers import AutoModelForImageSegmentation

from tqdm import tqdm

class ImageSegmenter:
    def __init__(self, model_path='models/RMBG-2.0'):
        """
        初始化图像分割器，加载模型

        Args:
            model_path (str): 模型的路径
        """
        self.model = AutoModelForImageSegmentation.from_pretrained(model_path, trust_remote_code=True)
        torch.set_float32_matmul_precision(['high', 'highest'][0])
        self.model.to('cuda')
        self.model.eval()

        # 数据设置
        self.image_size = (1024, 1024)
        self.transform_image = transforms.Compose([
            transforms.Resize(self.image_size),
            transforms.ToTensor(),
            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
        ])

    def process_image(self, input_image_path):
        """
        处理输入图像，生成掩膜并保存掩膜图像

        Args:
            input_image_path (str): 输入图像的路径

        Returns:
            PIL.Image.Image: 生成的掩膜图像
        """
        # 读取输入图像
        image = Image.open(input_image_path)
        input_images = self.transform_image(image).unsqueeze(0).to('cuda')

        # 预测
        with torch.no_grad():
            preds = self.model(input_images)[-1].sigmoid().cpu()
        pred = preds[0].squeeze()
        pred_pil = transforms.ToPILImage()(pred)
        mask = pred_pil.resize(image.size)

        # 保存掩膜图像
        mask_output_path = f"data/mask_image/mask-{os.path.splitext(os.path.basename(input_image_path))[0]}.png"
        mask.save(mask_output_path)  # 保存掩膜图像

        return mask
    
def mask_image_list(image_paths):
    """
    处理图像列表，返回掩膜列表

    Args:
        image_paths (list): 输入图像路径列表

    Returns:
        list: 生成的掩膜图像列表
    """
    segmenter = ImageSegmenter()  # 创建图像分割器实例
    masks = []

    for image_path in tqdm(image_paths):
        mask = segmenter.process_image(image_path)  # 处理每个图像
        masks.append(mask)  # 将掩膜添加到列表中

    return masks


# 使用示例
if __name__ == "__main__":
    model_path = 'models/RMBG-2.0'  # 模型路径
    input_image_path = "data/key_frame/frame_00000000.jpg"  # 输入图像路径
    mask_output_path = "rmbgmask_image.png"  # 掩膜图像保存路径

    segmenter = ImageSegmenter(model_path)  # 创建图像分割器实例
    mask_image = segmenter.process_image(input_image_path)  # 处理图像
    print(type(mask_image))