Develop Reference

Custom transformation on Subset of CIFAR10 - PyTorch

I am trying to create a custom transformation to part of the CIFAR10 data set which superimposing of an image over the dataset. I was able to download the data and divide it into subsets. Using the following code:
transform_train = transforms.Compose([
transforms.RandomCrop(32, padding=4),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])
traindata = datasets.CIFAR10('./data', train=True, download=True,
transform= transform_train)
partitions = 5
traindata_split = torch.utils.data.random_split(traindata, [int(traindata.data.shape[0] / partitions) for _ in range(partitions)])
then I wanted to modify part of the splits so I created the following class and functions to use as as follows:
class MyDataset(Dataset): # https://discuss.pytorch.org/t/torch-utils-data-dataset-random-split/32209/3
def __init__(self, subset, transform=None):
self.subset = subset
self.transform = transform
def __getitem__(self, index):
x, y = self.subset[index]
if self.transform:
x = self.transform(x)
return x, y
def __len__(self):
return len(self.subset)
and
class ImageSuperImpose(object):
""" Image input as PIL and output as PIL
To be used as part of torchvision.transforms
Args: p, a threshold value to control image thinning
"""
def __init__(self, p=0):
self.p = p
def __call__(self, image):
img = cv2.imread('img.jpg')
img = img('float32')/255
imgSm = cv2.resize(img,(32,32))
np_arr = image.cpu().detach().numpy().T
sample = cv2.addWeighted(np_arr, 1, imgSm, 1, 0)
sample = sample.T
t = torch.from_numpy(sample)
return sample
transform_train2 = transforms.Compose([
transforms.RandomCrop(32, padding=4),
transforms.RandomHorizontalFlip(),
ImagePoisoning(),
transforms.ToTensor(),
transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])
datasetA = MyDataset(
traindata_split[0], transform= transform_train2
)
test_loader = torch.utils.data.DataLoader(datasetA, batch_size=128, shuffle=True)
But when I tried to train the model on the subset I got the following error:
RuntimeError: The size of tensor a (32) must match the size of tensor b (3) at non-singleton dimension 0
** UPDATE**
Here is the full given error
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-20-7428084b03be> in <module>()
----> 1 train(model, opt, test_loader, 3)
9 frames
<ipython-input-14-fcb03e1d7685> in client_update(client_model, optimizer, train_loader, epoch)
5 client_model.train()
6 for e in range(epoch):
----> 7 for batch_idx, (data, target) in enumerate(train_loader):
8 data, target = data.to(device), target.to(device)
9 optimizer.zero_grad()
/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py in __next__(self)
433 if self._sampler_iter is None:
434 self._reset()
--> 435 data = self._next_data()
436 self._num_yielded += 1
437 if self._dataset_kind == _DatasetKind.Iterable and \
/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py in _next_data(self)
473 def _next_data(self):
474 index = self._next_index() # may raise StopIteration
--> 475 data = self._dataset_fetcher.fetch(index) # may raise StopIteration
476 if self._pin_memory:
477 data = _utils.pin_memory.pin_memory(data)
/usr/local/lib/python3.6/dist-packages/torch/utils/data/_utils/fetch.py in fetch(self, possibly_batched_index)
42 def fetch(self, possibly_batched_index):
43 if self.auto_collation:
---> 44 data = [self.dataset[idx] for idx in possibly_batched_index]
45 else:
46 data = self.dataset[possibly_batched_index]
/usr/local/lib/python3.6/dist-packages/torch/utils/data/_utils/fetch.py in <listcomp>(.0)
42 def fetch(self, possibly_batched_index):
43 if self.auto_collation:
---> 44 data = [self.dataset[idx] for idx in possibly_batched_index]
45 else:
46 data = self.dataset[possibly_batched_index]
<ipython-input-7-1bde43acaff0> in __getitem__(self, index)
7 x, y = self.subset[index]
8 if self.transform:
----> 9 x = self.transform(x)
10 return x, y
11
/usr/local/lib/python3.6/dist-packages/torchvision/transforms/transforms.py in __call__(self, img)
65 def __call__(self, img):
66 for t in self.transforms:
---> 67 img = t(img)
68 return img
69
/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
725 result = self._slow_forward(*input, **kwargs)
726 else:
--> 727 result = self.forward(*input, **kwargs)
728 for hook in itertools.chain(
729 _global_forward_hooks.values(),
/usr/local/lib/python3.6/dist-packages/torchvision/transforms/transforms.py in forward(self, tensor)
224 Tensor: Normalized Tensor image.
225 """
--> 226 return F.normalize(tensor, self.mean, self.std, self.inplace)
227
228 def __repr__(self):
/usr/local/lib/python3.6/dist-packages/torchvision/transforms/functional.py in normalize(tensor, mean, std, inplace)
282 if std.ndim == 1:
283 std = std.view(-1, 1, 1)
--> 284 tensor.sub_(mean).div_(std)
285 return tensor
286
RuntimeError: The size of tensor a (32) must match the size of tensor b (3) at non-singleton dimension 0

ValueError: tile cannot extend outside image pytorch

I am loading NumPy arrays as images in PyTorch while training the model, it’s giving me this error, I tried everything but couldn’t figure out pls help…, I am training a classifier model.......................................................................................................................................................................................
ValueError Traceback (most recent call last)
<ipython-input-12-b4d3f7be01c1> in <module>
1 # training
----> 2 trained_model = train(n_epochs, np.Inf, loaders, model, optimizer, criterion)
<ipython-input-10-b4d180a2c041> in train(n_epochs, valid_loss_min_input, loaders, model, optimizer, criterion, device, checkpoint_path, best_model_path)
29 ###################
30 model.train()
---> 31 for batch_idx, (data, target) in enumerate(loaders['train']):
32 # move to gpu
33 data, target = data.to(device), target.to(device)
G:\anaconda3\envs\data_env\lib\site-packages\torch\utils\data\dataloader.py in __next__(self)
343
344 def __next__(self):
--> 345 data = self._next_data()
346 self._num_yielded += 1
347 if self._dataset_kind == _DatasetKind.Iterable and \
G:\anaconda3\envs\data_env\lib\site-packages\torch\utils\data\dataloader.py in _next_data(self)
383 def _next_data(self):
384 index = self._next_index() # may raise StopIteration
--> 385 data = self._dataset_fetcher.fetch(index) # may raise StopIteration
386 if self._pin_memory:
387 data = _utils.pin_memory.pin_memory(data)
G:\anaconda3\envs\data_env\lib\site-packages\torch\utils\data\_utils\fetch.py in fetch(self, possibly_batched_index)
42 def fetch(self, possibly_batched_index):
43 if self.auto_collation:
---> 44 data = [self.dataset[idx] for idx in possibly_batched_index]
45 else:
46 data = self.dataset[possibly_batched_index]
G:\anaconda3\envs\data_env\lib\site-packages\torch\utils\data\_utils\fetch.py in <listcomp>(.0)
42 def fetch(self, possibly_batched_index):
43 if self.auto_collation:
---> 44 data = [self.dataset[idx] for idx in possibly_batched_index]
45 else:
46 data = self.dataset[possibly_batched_index]
G:\anaconda3\envs\data_env\lib\site-packages\torchvision\datasets\folder.py in __getitem__(self, index)
135 sample = self.loader(path)
136 if self.transform is not None:
--> 137 sample = self.transform(sample)
138 if self.target_transform is not None:
139 target = self.target_transform(target)
G:\anaconda3\envs\data_env\lib\site-packages\torchvision\transforms\transforms.py in __call__(self, img)
59 def __call__(self, img):
60 for t in self.transforms:
---> 61 img = t(img)
62 return img
63
<ipython-input-3-88cdee8f0d6c> in __call__(self, img)
7 im = np.asarray(img)
8 im = detect(im)
----> 9 img = Image.fromarray(im)
10 img = img.resize(size=(128, 128))
11 return img
G:\anaconda3\envs\data_env\lib\site-packages\PIL\Image.py in fromarray(obj, mode)
2768 obj = obj.tostring()
2769
-> 2770 return frombuffer(mode, size, obj, "raw", rawmode, 0, 1)
2771
2772
G:\anaconda3\envs\data_env\lib\site-packages\PIL\Image.py in frombuffer(mode, size, data, decoder_name, *args)
2708 return im
2709
-> 2710 return frombytes(mode, size, data, decoder_name, args)
2711
2712
G:\anaconda3\envs\data_env\lib\site-packages\PIL\Image.py in frombytes(mode, size, data, decoder_name, *args)
2648
2649 im = new(mode, size)
-> 2650 im.frombytes(data, decoder_name, args)
2651 return im
2652
G:\anaconda3\envs\data_env\lib\site-packages\PIL\Image.py in frombytes(self, data, decoder_name, *args)
795 # unpack data
796 d = _getdecoder(self.mode, decoder_name, args)
--> 797 d.setimage(self.im)
798 s = d.decode(data)
799
ValueError: tile cannot extend outside the image
In the custom function, I am trying to preprocess the image and then calling it
It works fine when I test out the code loading a small batch for display

Pytorch custom Dataset class giving wrong output

I am trying to use this class I built for a dataset but it saying that it should be a PIL or ndarray. Im not quite sure whats wrong with it. Here is the class that I am using
class RotateDataset(Dataset):
def __init__(self, image_list, size,transform = None):
self.image_list = image_list
self.size = size
self.transform = transform
def __len__(self):
return len(self.image_list)
def __getitem__(self, idx):
img = cv2.imread(self.image_list[idx])
image_height, image_width = img.shape[:2]
print("ID: ", idx)
if idx % 2 == 0:
label = 0 # Set label
# chose negative or positive rotation
rotation_degree = random.randrange(35, 50, 1)
posnegrot = np.random.randint(2)
if posnegrot == 0:
#positive rotation
#rotation_matrix = cv2.getRotationMatrix2D((num_cols/2, num_rows/2), rotation_degree, 1)
#img = cv2.warpAffine(img, rotation_matrix, (num_cols, num_rows))
img = rotate_image(img, rotation_degree)
img = crop_around_center(img, *largest_rotated_rect(image_width,
image_height,
math.radians(rotation_degree)))
else:
# Negative rotation
rotation_degree = -rotation_degree
img = crop_around_center(img, *largest_rotated_rect(image_width,
image_height,
math.radians(rotation_degree)))
else:
label = 1
img = cv2.resize(img, self.size, cv2.INTER_AREA)
return self.transform(img), self.transform(label)
The error that it is giving me is
TypeError: pic should be PIL Image or ndarray. Got class 'int'
It should give me a img (tensor) and a label (tensor)
but I dont think it is doing it correctly.
TypeError Traceback (most recent call last)
<ipython-input-34-f47943b2600c> in <module>
2 train_loss = 0.0
3 net.train()
----> 4 for image, label in enumerate(train_loader):
5 if train_on_gpu:
6 image, label = image.cuda(), label.cuda()
~\Anaconda3\envs\TF2\lib\site-packages\torch\utils\data\dataloader.py in __next__(self)
343
344 def __next__(self):
--> 345 data = self._next_data()
346 self._num_yielded += 1
347 if self._dataset_kind == _DatasetKind.Iterable and \
~\Anaconda3\envs\TF2\lib\site-packages\torch\utils\data\dataloader.py in _next_data(self)
383 def _next_data(self):
384 index = self._next_index() # may raise StopIteration
--> 385 data = self._dataset_fetcher.fetch(index) # may raise StopIteration
386 if self._pin_memory:
387 data = _utils.pin_memory.pin_memory(data)
~\Anaconda3\envs\TF2\lib\site-packages\torch\utils\data\_utils\fetch.py in fetch(self, possibly_batched_index)
42 def fetch(self, possibly_batched_index):
43 if self.auto_collation:
---> 44 data = [self.dataset[idx] for idx in possibly_batched_index]
45 else:
46 data = self.dataset[possibly_batched_index]
~\Anaconda3\envs\TF2\lib\site-packages\torch\utils\data\_utils\fetch.py in <listcomp>(.0)
42 def fetch(self, possibly_batched_index):
43 if self.auto_collation:
---> 44 data = [self.dataset[idx] for idx in possibly_batched_index]
45 else:
46 data = self.dataset[possibly_batched_index]
<ipython-input-28-6c77357ff619> in __getitem__(self, idx)
35 label = 1
36 img = cv2.resize(img, self.size, cv2.INTER_AREA)
---> 37 return self.transform(img), self.transform(label)
~\Anaconda3\envs\TF2\lib\site-packages\torchvision\transforms\transforms.py in __call__(self, pic)
99 Tensor: Converted image.
100 """
--> 101 return F.to_tensor(pic)
102
103 def __repr__(self):
~\Anaconda3\envs\TF2\lib\site-packages\torchvision\transforms\functional.py in to_tensor(pic)
53 """
54 if not(_is_pil_image(pic) or _is_numpy(pic)):
---> 55 raise TypeError('pic should be PIL Image or ndarray. Got {}'.format(type(pic)))
56
57 if _is_numpy(pic) and not _is_numpy_image(pic):
TypeError: pic should be PIL Image or ndarray. Got <class 'int'>

As discussed in the comments, the problem was applying transform on label as well. The label should instead simply be written as tensor:
return self.transform(img), torch.tensor(label)

Pytorch double image input : broken pipe with dataloader

I am developing a neural network in python with pytorch in order to classify a dataset of pairs of images. So I wanted to return as an output the 2 images and the ground truth but whenever I try to use the data loader I get the error "Broken pipe".
I want it to work as a classifier so I'm using this link (pytorch classifier CIFAR10) as an example
Here is my code :
###Defining class
class continuousImgDataset(Dataset):
def __init__(self, tabDataset, transform=None):
"""
Args:
tabDataset : Contains the imported data like : Item 1 = [im1,im2,ground-truth]
transform (callable, optional): Optional transform to be applied
on a sample.
"""
self.tabPairImg = tabDataset
self.transform = transform
def __len__(self):
return len(self.tabPairImg)
def __getitem__(self, idx):
img1 = self.tabPairImg[idx][0]
img2 = self.tabPairImg[idx][1]
label = self.tabPairImg[idx][2]
if self.transform:
img1 = self.transform(img1)
img2 = self.transform(img2)
return img1, img2,label
transform = transforms.Compose(
[transforms.Scale((32,32)),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
###Make the training and testing dataset
custom_dataset = continuousImgDataset(pairImg,transform)
train_dataset = continuousImgDataset(pairImg[: int(len(custom_dataset) * .90)],transform)
test_dataset = continuousImgDataset(pairImg[int(len(custom_dataset) * .90) : int(len(custom_dataset))],transform)
print(len(custom_dataset.tabPairImg)) #Output :22620
print(len(train_dataset.tabPairImg)) #Output : 20358
print(len(test_dataset.tabPairImg)) #Output : 2262
### Loaders
trainloader = torch.utils.data.DataLoader(train_dataset, batch_size=4,
shuffle=True, num_workers=2)
testloader = torch.utils.data.DataLoader(test_dataset, batch_size=4,
shuffle=False, num_workers=2)
classes = ('joint','disjoint')
### Image show and Error
"""From the official website but a little modified to have 2 imgs"""
def imshow(img):
img = img / 2 + 0.5 # unnormalize
npimg = img.numpy()
plt.imshow(np.transpose(npimg, (1, 2, 0)))
plt.show()
# get some random training images
dataiter = iter(trainloader) #ERROR HERE : Broken pipe
images1,images2, labels = dataiter.next()
# show images
imshow(torchvision.utils.make_grid(images1))
imshow(torchvision.utils.make_grid(images2))
# print labels
print(' '.join('%5s' % classes[labels[j]] for j in range(2)))
I may not have fully understand what the dataloader was expecting in order to iterate on it. Any help would be appreciated :)
EDIT : Here is the full error :
BrokenPipeError Traceback (most recent call last)
<ipython-input-58-314e85cc8fbb> in <module>
7
8 # get some random training images
----> 9 dataiter = iter(trainloader)
10 images1,images2, labels = dataiter.next()
11
C:\ProgramData\Anaconda3\lib\site-packages\torch\utils\data\dataloader.py in __iter__(self)
191
192 def __iter__(self):
--> 193 return _DataLoaderIter(self)
194
195 def __len__(self):
C:\ProgramData\Anaconda3\lib\site-packages\torch\utils\data\dataloader.py in __init__(self, loader)
467 # before it starts, and __del__ tries to join but will get:
468 # AssertionError: can only join a started process.
--> 469 w.start()
470 self.index_queues.append(index_queue)
471 self.workers.append(w)
C:\ProgramData\Anaconda3\lib\multiprocessing\process.py in start(self)
110 'daemonic processes are not allowed to have children'
111 _cleanup()
--> 112 self._popen = self._Popen(self)
113 self._sentinel = self._popen.sentinel
114 # Avoid a refcycle if the target function holds an indirect
C:\ProgramData\Anaconda3\lib\multiprocessing\context.py in _Popen(process_obj)
221 #staticmethod
222 def _Popen(process_obj):
--> 223 return _default_context.get_context().Process._Popen(process_obj)
224
225 class DefaultContext(BaseContext):
C:\ProgramData\Anaconda3\lib\multiprocessing\context.py in _Popen(process_obj)
320 def _Popen(process_obj):
321 from .popen_spawn_win32 import Popen
--> 322 return Popen(process_obj)
323
324 class SpawnContext(BaseContext):
C:\ProgramData\Anaconda3\lib\multiprocessing\popen_spawn_win32.py in __init__(self, process_obj)
87 try:
88 reduction.dump(prep_data, to_child)
---> 89 reduction.dump(process_obj, to_child)
90 finally:
91 set_spawning_popen(None)
C:\ProgramData\Anaconda3\lib\multiprocessing\reduction.py in dump(obj, file, protocol)
58 def dump(obj, file, protocol=None):
59 '''Replacement for pickle.dump() using ForkingPickler.'''
---> 60 ForkingPickler(file, protocol).dump(obj)
61
62 #
BrokenPipeError: [Errno 32] Broken pipe

How to resolve Type error - Lime interpret CNN results

I am using keras with Tesorflow backend (using python 2.7) and I built two dense layer on top of pretrained vgg16 model with four classes. And I get a pretty decent results on my validation set. Now I want to use lime to interpret my result.
I import the lime package and transform one of my images following the lime github repo https://github.com/marcotcr/lime/blob/master/doc/notebooks/Tutorial%20-%20Image%20Classification%20Keras.ipynb. My path_list contains one photo.
import lime
from lime import lime_image
def transform_img_fn(path_list):
out = []
for img_path in path_list:
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img) / 255
x = np.expand_dims(x, axis=0)
out.append(x)
return np.vstack(out)
check_image = transform_img_fn(path_list)
Then
check_image[0].shape
OUTPUT: (3, 224, 224)
predictions[0]
OUTPUT: array([9.67346e-01, 3.00240e-03, 2.96037e-02, 4.79915e-05], dtype=float32)
explainer = lime_image.LimeImageExplainer()
explanation = explainer.explain_instance(check_image[0], model_top.predict, hide_color=0, num_samples=100)
I get this error:
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-115-e286e97e849d> in <module>()
----> 1 explanation = explainer.explain_instance(check_image[0], model_top.predict, hide_color=0, num_samples=100)
/home/ec2-user/anaconda2/envs/env1/lib/python2.7/site-packages/lime/lime_image.pyc in explain_instance(self, image, classifier_fn, labels, hide_color, top_labels, num_features, num_samples, batch_size, segmentation_fn, distance_metric, model_regressor, random_seed)
165 segmentation_fn = SegmentationAlgorithm('quickshift', kernel_size=4,
166 max_dist=200, ratio=0.2,
--> 167 random_seed=random_seed)
168 try:
169 segments = segmentation_fn(image)
/home/ec2-user/anaconda2/envs/env1/lib/python2.7/site-packages/lime/wrappers/scikit_image.pyc in __init__(self, algo_type, **target_params)
103 if (self.algo_type == 'quickshift'):
104 BaseWrapper.__init__(self, quickshift, **target_params)
--> 105 kwargs = self.filter_params(quickshift)
106 self.set_params(**kwargs)
107 elif (self.algo_type == 'felzenszwalb'):
/home/ec2-user/anaconda2/envs/env1/lib/python2.7/site-packages/lime/wrappers/scikit_image.pyc in filter_params(self, fn, override)
82 result = {}
83 for name, value in self.target_params.items():
---> 84 if has_arg(fn, name):
85 result.update({name: value})
86 result.update(override)
/home/ec2-user/anaconda2/envs/env1/lib/python2.7/site-packages/lime/utils/generic_utils.pyc in has_arg(fn, arg_name)
19 else:
20 try:
---> 21 arg_spec = inspect.getargspec(fn.__call__)
22 except AttributeError:
23 return False
/home/ec2-user/anaconda2/envs/env1/lib/python2.7/inspect.pyc in getargspec(func)
813 func = func.im_func
814 if not isfunction(func):
--> 815 raise TypeError('{!r} is not a Python function'.format(func))
816 args, varargs, varkw = getargs(func.func_code)
817 return ArgSpec(args, varargs, varkw, func.func_defaults)
TypeError: <method-wrapper '__call__' of builtin_function_or_method object at 0x7fea20ea4e60> is not a Python function
Based on the documentation, "classifier_fn: function that takes a list of images and returns a matrix of prediction probabilities". I replaced this argument with model_top.predict. I can get all of my predictions if I call predictions = model_top.predict(validation_data, batch_size=32)
Any help would be appreciated.