NFT Top1000 Visual-Text Dataset

NFT Top1000 Visual-Text Dataset是由一组研究者在计算机视觉领域构建的首个NFT相关视觉-文本数据集，包含7.56百万图像-文本对，来源于以太坊区块链上销售量最高的1000个PFP NFT集合。数据集创建过程中，研究者从智能合约中获取资源链接，下载并筛选出具有实际语义信息的NFT项目。该数据集适用于NFT领域的检索、生成和视觉问答等下游任务，旨在解决大规模、高相似度图像-文本检索的挑战。

The NFT Top1000 Visual-Text Dataset is the first NFT-related visual-text dataset constructed by a group of computer vision researchers. It comprises 7.56 million image-text pairs sourced from the top 1000 PFP NFT collections ranked by trading volume on the Ethereum blockchain. During the dataset construction process, researchers obtained resource links from smart contracts, downloaded and filtered NFT projects with actual semantic information. This dataset is applicable to downstream tasks in the NFT domain such as retrieval, generation, and visual question answering, aiming to address the challenges of large-scale, high-similarity image-text retrieval.