xpark.dataset.ImageWatermarkDetect

class xpark.dataset.ImageWatermarkDetect(_local_model: str = 'finetrainers/laion-watermark-detection')

Image watermark detection processor for CPU, GPU.

Runs the watermark classifier over each input image and returns the predicted probability that the image contains a watermark, in [0, 1]. Higher scores mean the image is more likely to be watermarked.

Parameters:

_local_model – The watermark detection model name for CPU or GPU. default is “finetrainers/laion-watermark-detection”. available models: [‘finetrainers/laion-watermark-detection’]

Examples

from xpark.dataset.expressions import col
from xpark.dataset import ImageWatermarkDetect, from_items
import numpy as np

ds = from_items([
    {"image": np.random.randint(0, 255, (256, 256, 3)).astype(np.uint8), "path": "test.jpg"}
])
ds = ds.with_column(
    "image_watermark_prob",
    ImageWatermarkDetect().options(num_workers={"CPU": 4}, batch_size=32).with_column(col("image")),
)
print(ds.take(1))

Methods

__call__(images)	Call self as a function.
options(**kwargs)
with_column(images)

__call__(images: pa.ChunkedArray) → pa.Array

Call self as a function.

options(**kwargs: Unpack[ExprUDFOptions]) → Self

with_column(images: pa.ChunkedArray) → pa.Array