"IsADirectoryError: [Errno 21] Is a directory: " It is a file

Question

I already split the data into test and training set into the different folder. Now I need to load the patient data. Each patient has 8 images.

def load_dataset(root_dir, split):
"""
load the data set numpy arrays saved by the preprocessing script
:param root_dir: path to input data
:param split: defines whether to load the training or test set
:return: data: dictionary containing one dictionary ({'data', 'seg', 'pid'}) per patient
"""
in_dir = os.path.join(root_dir, split)
data_paths = [os.path.join(in_dir, f) for f in os.listdir(in_dir)]
data_and_seg_arr = [np.load(ii, mmap_mode='r') for ii in data_paths]
pids = [ii.split('/')[-1].split('.')[0] for ii in data_paths]
data = OrderedDict()
for ix, pid in enumerate(pids):
    data[pid] = {'data': data_and_seg_arr[ix][..., 0], 'seg': data_and_seg_arr[ix][..., 1], 'pid': pid}
return data

But, the error said:

File "/home/zhe/Research/Seg/heart_seg/data_loader.py", line 61, in load_dataset
data_and_seg_arr = [np.load(ii, mmap_mode='r') for ii in data_paths]
File "/home/zhe/Research/Seg/heart_seg/data_loader.py", line 61, in <listcomp>
data_and_seg_arr = [np.load(ii, mmap_mode='r') for ii in data_paths]
File "/home/zhe/anaconda3/envs/tf_env/lib/python3.6/site-packages/numpy/lib/npyio.py", line 372, in load
fid = open(file, "rb")
IsADirectoryError: [Errno 21] Is a directory: './data/preprocessed_data/train/Patient009969'

It is already a file name, not a directory. Thanks!

Answer 1

It seems that ./data/preprocessed_data/train/Patient009969 is a directory, not a file.

os.listdir() returns both files and directories.

Maybe try using os.walk() instead. It treats files and directories separately, and can recurse inside the subdirectories to find more files in a iterative way:

data_paths = [os.path.join(pth, f) 
    for pth, dirs, files in os.walk(in_dir) for f in files]

Answer 2

Do you have both files and directories inside your path? os.listdir will list both files and directories, so when you try to open a directory with np.load it will give that error. You can filter only files to avoid the error:

data_paths = [os.path.join(in_dir, f) for f in os.listdir(in_dir)]
data_paths = [i for i in data_paths if os.path.isfile(i)]

Or all together in a single line:

data_paths = [i for i in (os.path.join(in_dir, f) for f in os.listdir(in_dir)) if os.path.isfile(i)]

Answer 3

I had the same problem but i resolved by changing my path from Data/Train_Data/myDataset/(my images) to Data/Train_Data/(my images) where the script python is in the same path as Data. Hope this help.