TypeError: img should be PIL Image. Got <class 'numpy.ndarray'>

1

Train_loc = ‘./Mehsana/Train_data/’
Val_loc = ‘./Mehsana/Val_data/’

X = glob.glob(Train_loc+‘Images/’+‘img’)
Y = glob.glob(Train_loc +‘labels/’+’.img’)

X_train, Y_train, X_test, Y_test = X[:2000], Y[:2000], X[2000:], Y[2000:]

del X,Y

class CustomDataset(Dataset):
def init(self, image_paths, target_paths, train=True): # initial logic happens like transform

    self.image_paths = image_paths
    self.target_paths = target_paths
   
    self.transforms  = transforms.Compose([
#transforms.RandomHorizontalFlip(p=0.5),
#transforms.RandomVerticalFlip(p=0.5), 
transforms.Resize((224,224)), transforms.ToTensor(),
transforms.Normalize(

mean= [688.8383261571815, 890.7605253921796, 1012.9213980828494, 3228.524152857917],
std=[238.88659552814724, 267.57694692565684, 347.344288531034, 689.5044531578629])])

def __getitem__(self, index):

    img =gdal.Open(self.image_paths[index],gdal.GA_ReadOnly)
    img= img.ReadAsArray()
    print(img.shape)
    img = np.moveaxis(img,0,-1)
 
    mask = gdal.Open(self.target_paths[index],gdal.GA_ReadOnly)
    print(np.unique(mask))
    t_image = self.transforms(img)
    return t_image, mask

def __len__(self):  # return count of sample we have

    return len(self.image_paths)

train_dataset = CustomDataset( X_train, Y_train, train=True)

train_dataset = CustomDataset(train_image_paths, train_mask_paths, train=True)

I get this error when I am trying to generate a custom dataset in pytorch for Unet model. My images are of size (250,250,4).

1 Like
2

You need to add transfroms.ToTensor() at the end of transformations block and custom DataLoaders require operations with PIL library. So, you can do transformations with numpy and then convert the image to PIL format. In your case, you need to convert the result of img = np.moveaxis(img,0,-1) to PIL and then pass it to transform function.

1 Like