模型动物园

模型动物园是一个包含30个模型的集合，这些模型由亚马逊网络服务训练，用于不同领域的图像分类任务。数据集涵盖了从25K到1.8M训练图像的大型公共数据集，如NWPU-RESISC45、Food-101等，旨在通过提供多样化的预训练模型来增强对新目标任务的微调效果。模型动物园的应用领域包括提高在数据有限情况下的测试准确性，通过模型选择算法快速找到适合特定任务的预训练模型，从而避免耗时的暴力微调过程。

Model Zoo is a collection of 30 models trained by Amazon Web Services (AWS) for image classification tasks across various domains. These models are developed based on large-scale public datasets with training image quantities ranging from 25,000 to 1.8 million, including NWPU-RESISC45, Food-101, and other similar datasets. The core goal of this Model Zoo is to boost the fine-tuning performance on new target tasks by offering a diverse set of pre-trained models. Its application scenarios include improving test accuracy when training data is limited, enabling rapid selection of pre-trained models suitable for specific tasks via model selection algorithms, thus avoiding time-consuming brute-force fine-tuning procedures.