xkplsz 发表于 2024-2-17 20:37:47

基于 Pytorch 的AI 去水印方法 Watermark-Removal-Pytorch github 还有哪些好用的方法?

Watermark Removal using Deep Image Priors with Pytorch
https://github.com/braindotai/Watermark-Removal-Pytorch



适用于一定情景,一定的效果. 没有PS那么自由强大,但是可以代码直接使用.不用开PS.



lpzyuan 发表于 2024-7-29 12:13:44

{:1_236:}

发表于 1970-1-1 08:00:00

发表于 1970-1-1 08:00:00

发表于 1970-1-1 08:00:00

发表于 1970-1-1 08:00:00

LLLLhegxuE 发表于 2024-2-17 21:22:24

https://quququ.cn/


基于AI的 收费的 ,估计算法就是 楼上的PyTorch 的
但是做好界面,包装成网页这个是真的有门槛的, 所以高清的要收点小钱.




LLLLhegxuE 发表于 2024-2-17 21:16:26


PS 各种水印去除, 如果能自动就这些技能就全费了
图片水印的类型很多. 很难用一招吃遍.
https://www.bilibili.com/video/BV1qu411t7CG/?spm_id_from=333.337.search-card.all.click&vd_source=ae1d5f8d0c7abdd93954ea3bcebcb863




UEBoy-Scout 发表于 2024-2-17 20:57:58


bing_watermark_remover_pythonPublic
没有万能去水印方法.
GIT 其他方法 - 直到特定区域 的 CV2 方法.

bing_watermark_remover_python/app.py at master · AmirHosseinAsa/bing_watermark_remover_python · GitHub

import cv2
import numpy as np
import os

def remove_watermark(image_path, output_path):
    # Load the image
    image = cv2.imread(image_path)

    # Determine the coordinates of the region to remove
    bottom_left_x = 0
    bottom_left_y = image.shape - 50
    top_right_x = 50
    top_right_y = image.shape

    # Create a mask for the main region to remove
    mask = np.zeros(image.shape[:2], dtype=np.uint8)
    mask = 255

    # Pre-process the image (optional)
    # Apply image denoising or blurring techniques

    # Inpaint the main region
    image_without_watermark = cv2.inpaint(image, mask, 3, cv2.INPAINT_TELEA)

    # Create a mask for the remaining small region to remove
    small_mask = np.zeros(image.shape[:2], dtype=np.uint8)
    small_mask[-60:, :15] = 255

    # Inpaint the remaining small region
    image_without_watermark = cv2.inpaint(image_without_watermark, small_mask, 3, cv2.INPAINT_TELEA)

    # Post-process the image (optional)
    # Apply additional adjustments or refinements, such as blending or texture synthesis

    # Save the image without the watermark
    cv2.imwrite(output_path, image_without_watermark)

# Input and output folders
input_folder = "input"
output_folder = "output"

# Iterate over the files in the input folder
for filename in os.listdir(input_folder):
    if filename.endswith(".jpeg") or filename.endswith(".jpg"):
      # Create the input and output paths for each image
      image_path = os.path.join(input_folder, filename)
      output_path = os.path.join(output_folder, filename)

      # Remove watermark and save the image
      remove_watermark(image_path, output_path)

发表于 1970-1-1 08:00:00

页: [1]
查看完整版本: 基于 Pytorch 的AI 去水印方法 Watermark-Removal-Pytorch github 还有哪些好用的方法?