Xiaojie0903/Genview_syntheric_dataset_in1k
收藏Hugging Face2024-12-18 更新2024-12-21 收录
下载链接：
https://hf-mirror.com/datasets/Xiaojie0903/Genview_syntheric_dataset_in1k
下载链接
链接失效反馈官方服务：
资源简介：
GenView ImageNet数据集专为自监督学习设计，旨在通过自适应视图生成提升视图质量。该数据集使用预训练的CLIP ViT-H/14模型，通过PCA分析分离前景和背景，并根据前景比例调整噪声水平，确保生成的图像变体既具有多样性又保持语义一致性。
The GenView ImageNet Dataset is designed for self-supervised learning, focusing on enhancing view quality through adaptive view generation. It uses the pretrained CLIP ViT-H/14 backbone to create semantically consistent image variations by separating foreground and background using PCA analysis. Noise levels are adjusted based on the foreground ratio, ensuring both diversity and semantic consistency. The dataset is sourced from ImageNet and provides detailed steps for downloading and extracting the dataset.
提供机构：
Xiaojie0903
