ImageNet-21K
收藏资源简介:
ImageNet-21K数据集由穆罕默德·本·扎耶德人工智能大学创建,旨在通过生成一个较小但具有代表性的子集,从大型数据集中提取关键信息,以提高模型训练效率。该数据集包含20个IPC,通过链接可访问。数据集创建过程中采用了课程数据增强(CDA)技术,有效提升了合成数据的准确性。该数据集主要应用于解决大规模数据集的存储和计算挑战,支持资源有限的科研人员参与前沿模型训练和应用开发,同时有助于缓解数据隐私问题。
The ImageNet-21K dataset was developed by Mohamed bin Zayed University of Artificial Intelligence. It aims to extract critical information from large-scale datasets by generating a small yet representative subset, thereby improving model training efficiency. The dataset contains 20 IPC categories and is accessible via links. Course Data Augmentation (CDA) technology was adopted during the dataset creation process, which effectively enhanced the accuracy of synthetic data. This dataset is primarily used to address the storage and computational challenges of large-scale datasets, enabling researchers with limited resources to participate in cutting-edge model training and application development, while also helping to mitigate data privacy issues.

- 1Dataset Distillation in Large Data Era穆罕默德·本·扎耶德人工智能大学 · 2023年



