Hi,
For my problem, I have a dataset that contains train.csv with image names and labels + folder ‘train’ with images.
I have a separate dataset which I want to use as validation data. This dataset has val.csv and ‘val’ folder with images.
I’m not sure how to create databunch for training.
train_df = pd.read_csv(‘train.csv’)
path = Path(‘path-to-train/’)
path2 = Path(‘path-to-val/’)
train = [ImageList.from_df(df, path=path, cols=‘image’, folder=folder, suffix=’.jpeg’)
for df, folder in zip([train_df], [‘train’])]
tfms = get_transforms(do_flip=True)
data = (train[0].split_none()
.label_from_df(label_cls = FloatList)
.transform(tfms, size=224)
.databunch(path=Path(’.’), bs=64).normalize(imagenet_stats))