Midjourney V5.1 Cleaned Data

This is a cleaned subset of the larger, MidJourney scraped dataset (link to full https://www.kaggle.com/datasets/iraklip/midjourney-v51-prompts-and-image-links)

read more about the series at https://followfoxai.substack.com/p/releasing-vodka-v2-and-all-the-details

The main difference is that it contains only upscaled images as the main indicator of users' top choice among the initially generated ones, hopefully filtering out higher-quality generation.

Some additional cleaning was done to separate prompts from added metadata, as well a few parameter inputs were separated out (version parameter, aspect ratio)

本数据集为更大规模的中途之旅（MidJourney）抓取数据集（完整链接：https://www.kaggle.com/datasets/iraklip/midjourney-v51-prompts-and-image-links）的清洗子集。有关该系列更详细的信息，请参阅：https://followfoxai.substack.com/p/releasing-vodka-v2-and-all-the-details。其主要差异在于，它仅包含用户在最初生成图像中选择率最高的放大图像，以期筛选出更高品质的生成内容。此外，为了区分提示与附加元数据，以及分离出一些参数输入（如版本参数、宽高比），我们还进行了额外的清洗工作。