Watermark-or-Not-20K Dataset

Overview

The Watermark-or-Not-20K dataset consists of 20,000 images annotated with binary labels indicating the presence or absence of a watermark. It is designed to support training and evaluation of models focused on watermark detection, which is useful for content filtering, copyright protection, and image moderation tasks.

Dataset Structure

• Split: train
• Number of samples: 20,000
• Label Type: Categorical (2 classes)
• Image Resolution: Ranges from 158 pixels to 4.93k pixels in width
• Storage Format: Auto-converted to Parquet for efficient access

Label Classes

The dataset contains the following classes:

• 0 - No Watermark
• 1 - Watermark

Usage

The dataset can be accessed using the Hugging Face datasets library:

from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Watermark-or-Not-20K")

Applications

This dataset is suitable for:

• Training computer vision models to detect watermarks
• Fine-tuning transformer-based vision models on binary classification tasks
• Building AI-based content moderation pipelines