根据原文的描述:Based on the ImageNet dataset, we propose the ImageNet-S dataset with 1.2 million training images and 50k high-quality semantic segmentation annotations for evaluation。
🌷🍁 博主猫头虎(🐅🐾)带您 Go to New World✨🍁 🦄 博客首页——🐅🐾猫头虎的博客🎐 🐳 《面试题大全专栏》 🦕 文章图文…