TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Show HN: We collected detailed annotations for text-to-image generation

2 点作者 maalber4 个月前
Recently, the most popular modality text-to-image annotations has been preference data, where annotators usually choose between two images two indicate their favorite. While this does work to fine-tune models, it lacks additional information about what might be wrong with the images. E.g., what part of the image is misaligned relative to the prompt. Google research propose a modality for more information rich annotations (<a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2312.10240" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2312.10240</a>). Based on this, we produced this dataset of ~13k images. We collected in total ~1.5 million annotations from 150k annotators using our annotation API. If you are interested you can learn more about the API at <a href="https:&#x2F;&#x2F;docs.rapidata.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;docs.rapidata.ai&#x2F;</a><p>Let me know if you have any questions about the dataset or Rapidata in general!

暂无评论

暂无评论