I have a collection of images whose file names has an alphabet, eg, “234e7.jpg” and “123a.jpg”. I am able to create a datablock as below and even view the images (with labels) using show_batch() without issues
Birds = DataBlock(blocks=(ImageBlock, MultiCategoryBlock),
splitter=ColSplitter(),
get_x=lambda x:image_source/“train”/f’{x[0]}’,
get_y=lambda x:x[1].split(’ '),
item_tfms=Resize(224),
batch_tfms=aug_transforms())
dls = Birds.dataloaders(df_train)
dls.show_batch(max_n=2, figsize=(10,10))
When I call fit_one_cycle, I get an issue where ‘e’ is present in the filename as the it is interpreted like scientific notation:
learn = cnn_learner(dls, resnet34, pretrained=True, metrics=error_rate).to_fp16()
learn.fit_one_cycle(4)
Error:
"FileNotFoundError: [Errno 2] No such file or directory: ‘/./train/2340000000.jpg’
I want the “learn” object to read the file name as “234e7.jpg” and not as “2340000000.jpg”. Any ideas?