Unverified50% confidenceFactExact time
CLIP的训练数据来自互联网公开爬取的数据集WIT（WebImageText），其规模约为ImageNet（约1400万张标注图像）的近30倍
1
Sources
50%
Confidence
Long-term
Relevance
8/6/2026
First Seen
Sources
Related Claims
UnverifiedCommon Crawl项目存储了超过2500亿网页的完整HTML快照，压缩后数据量达PB级别，是许多AI模型的主要训练数据来源之一78% similarUnverifiedThe training data for modern image generation models typically comes from billions of internet images and their accompanying text descriptions.74% similarUnverifiedCLIP was released in 2021 and trained on 400 million image-text paired data collected from the internet.73% similarUnverifiedImageNet数据集花费数年、动用数万名众包标注员才完成140万张图像的分类标注65% similarUnverifiedWhisper's training data contains 680,000 hours, largely sourced from videos on the internet with varying subtitle quality.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/693254API
curl https://kongchang.com/api/v1/knowledge/claims/693254MCP
get_claim(id=693254)