spaicom-lab/semasia-tiny-imagenet

该数据集包含Tiny ImageNet数据集的潜在特征（latents），这些特征是通过多种预训练视觉模型（如AIM、BEiT、Cait、CoatNet、ConvFormer等）生成的。数据集以parquet格式存储，涵盖训练集和验证集分割，并支持多种模型配置和图像分辨率（如224x224、336x336、448x448等）。这些潜在特征可用于迁移学习、特征提取或模型分析等任务。This dataset contains latent features for the Tiny ImageNet dataset, generated by various pre-trained vision models (e.g., AIM, BEiT, Cait, CoatNet, ConvFormer). The data is stored in parquet format, includes train and validation splits, and supports multiple model configurations and image resolutions (e.g., 224x224, 336x336, 448x448). These latent features can be used for tasks such as transfer learning, feature extraction, or model analysis.