请问datasets文件夹里base.py文件里,img2_path为什么是随机读取的

[Tisword]

感觉怎么运行这个文件的步骤不是很详细，数据集也很难对应起来，请问能提供一份用于生成预训练的数据集，目录能和FineGRP.py读取train1和train2对应上的吗，试了下直接下载FineGRP数据集和这边对应不上，非常感谢！

[feiyu-flyfish]

Img2_path denotes random selectiong of samples in dataset2 to create the sample space, while the train1_dir and train2_dir are pretraining datasets 1 and 2 respectively.

[Tisword]

thank you ！Could you provide pretraining datasets 1 and 2 ？It's really helpful for me

[JeremyXSC]

Dataset1 denotes pre-trained dataset FineGPR, while Dataset2 represents target dataset for validation, e.g. Market-1501.