search

OpenGVLab / InternImage

[CVPR 2023 Highlight] InternImage: Exploring Large-Scale Vision Foundation Models with Deformable Convolutions
https://arxiv.org/abs/2211.05778
MIT License
2.48k stars 231 forks source link

同样的问题缺失val.txt #163

Closed orzlh closed 1 year ago

orzlh commented 1 year ago

您好，与之前的问题一样，我也采用了classification里第一种数据集方式，出现了同样的缺失问题。 那么，我如果想继续用第一种方式，需要做出哪些改变？还是说，必须更改我的数据集形式。

orzlh commented 1 year ago

我找到解决办法了，在classification/dataset/build.py中将L210-L216替换

            dataset = CachedImageFolder(root, transform=transform)
        elif prefix == 'val':
            root = os.path.join(config.DATA.DATA_PATH, 'val')
            dataset = CachedImageFolder(root, transform=transform)
jeonga0303 commented 3 months ago

Your solution was helpful.

tip from .cached_image_folder import ImageCephDataset, CachedImageFolder