license: bigscience-openrail-m
language:
- en
pretty_name: scarlet-dark
Overview
The 8chan initiative emanates from the Sleeping AI Lab, aiming to furnish the research community with a superlative dataset in unconventional domains. Aligned with Sleeping AI's Datasets initiative, our commitment is directed toward furnishing premium Open Source datasets, with 8chan standing out as a singular endeavour that delves into a segment of the Internet colloquially known as the Dark Web, often deemed taboo in mainstream media. Our contribution encompasses media and image data extracted from the 8chan image board, serving as the enigmatic counterpart to the 4chan platform. This pioneering dataset distinguishes itself within the research community, being the inaugural compilation exclusively dedicated to 8chan. We trust that the community will exercise due responsibility in its utilisation.
Technical Aspects of the Dataset
Within this dataset, designated as kun vartaint, individuals may discover unaltered media and images sourced from the entirety of the 8chan platform, encompassing all 488 boards and their respective threads up until November 30, 2023. Subsequently, these data were downloaded and stored within the overarching super-folder 'kun,' with each board's collected data residing in its distinct folder identified by the nomenclature "board = foldername." Metadata pertaining to the uploaded user accompanies the media, inclusive of sensitive information. Accordingly, we implore researchers to employ this information ethically and responsibly. Despite the presence of numerous corrupted images and media, their discerning examination reveals commensurate value. For practical engagement with the images, an image-specific subset of the scraped data embedding has been disseminated on Kaggle, fostering the development of robust and innovative models. Subsequently, we introduced a refined variant, clear_kun, representing a preprocessed and sanitised iteration of the dataset. This version, comprising 3882 images, has undergone meticulous filtering to eliminate corruptions and serves as the foundation for generating embeddings.
Liability
It is imperative to clarify that any potential misuse by third parties absolves the undersigned of responsibility. We uphold a stringent request policy, necessitating interested parties to submit requests for dataset access, which will be individually reviewed. Researchers are strongly encouraged to uphold privacy and adhere to ethical guidelines, with any inadvertent misuse falling outside the purview of responsibility. The release of this dataset is expressly intended for academic and research purposes, encompassing content should be viewed by 20 and above older individuals. For inquiries or concerns, please direct correspondence to [email protected].
Kaggle (Image Embedding): https://www.kaggle.com/datasets/sleepingcat4/8chan-image-embeddings/data