Chapter 27
Image Classification
NotebookPython 3 (ipykernel)73 cells
In [ ]python · cell 1
python
#hide
! [ -e /content ] && pip install -Uqq fastbook
import fastbook
fastbook.setup_book()In [ ]python · cell 2
python
#hide
from fastbook import *Image Classification
From Dogs and Cats to Pet Breeds
In [ ]python · cell 5
python
from fastai.vision.all import *
path = untar_data(URLs.PETS)In [ ]python · cell 6
python
#hide
Path.BASE_PATH = pathIn [ ]python · cell 7
python
path.ls()In [ ]python · cell 8
python
(path/"images").ls()In [ ]python · cell 9
python
fname = (path/"images").ls()[0]In [ ]python · cell 10
python
re.findall(r'(.+)_\d+.jpg$', fname.name)In [ ]python · cell 11
python
pets = DataBlock(blocks = (ImageBlock, CategoryBlock),
get_items=get_image_files,
splitter=RandomSplitter(seed=42),
get_y=using_attr(RegexLabeller(r'(.+)_\d+.jpg$'), 'name'),
item_tfms=Resize(460),
batch_tfms=aug_transforms(size=224, min_scale=0.75))
dls = pets.dataloaders(path/"images")Presizing
In [ ]python · cell 13
python
dblock1 = DataBlock(blocks=(ImageBlock(), CategoryBlock()),
get_y=parent_label,
item_tfms=Resize(460))
# Place an image in the 'images/grizzly.jpg' subfolder where this notebook is located before running this
dls1 = dblock1.dataloaders([(Path.cwd()/'images'/'grizzly.jpg')]*100, bs=8)
dls1.train.get_idxs = lambda: Inf.ones
x,y = dls1.valid.one_batch()
_,axs = subplots(1, 2)
x1 = TensorImage(x.clone())
x1 = x1.affine_coord(sz=224)
x1 = x1.rotate(draw=30, p=1.)
x1 = x1.zoom(draw=1.2, p=1.)
x1 = x1.warp(draw_x=-0.2, draw_y=0.2, p=1.)
tfms = setup_aug_tfms([Rotate(draw=30, p=1, size=224), Zoom(draw=1.2, p=1., size=224),
Warp(draw_x=-0.2, draw_y=0.2, p=1., size=224)])
x = Pipeline(tfms)(x)
#x.affine_coord(coord_tfm=coord_tfm, sz=size, mode=mode, pad_mode=pad_mode)
TensorImage(x[0]).show(ctx=axs[0])
TensorImage(x1[0]).show(ctx=axs[1]);Checking and Debugging a DataBlock
In [ ]python · cell 15
python
dls.show_batch(nrows=1, ncols=3)In [ ]python · cell 16
python
pets1 = DataBlock(blocks = (ImageBlock, CategoryBlock),
get_items=get_image_files,
splitter=RandomSplitter(seed=42),
get_y=using_attr(RegexLabeller(r'(.+)_\d+.jpg$'), 'name'))
pets1.summary(path/"images")In [ ]python · cell 17
python
learn = vision_learner(dls, resnet34, metrics=error_rate)
learn.fine_tune(2)Cross-Entropy Loss
Viewing Activations and Labels
In [ ]python · cell 20
python
x,y = dls.one_batch()In [ ]python · cell 21
python
yIn [ ]python · cell 22
python
preds,_ = learn.get_preds(dl=[(x,y)])
preds[0]In [ ]python · cell 23
python
len(preds[0]),preds[0].sum()Softmax
In [ ]python · cell 25
python
plot_function(torch.sigmoid, min=-4,max=4)In [ ]python · cell 26
python
#hide
torch.random.manual_seed(42);In [ ]python · cell 27
python
acts = torch.randn((6,2))*2
actsIn [ ]python · cell 28
python
acts.sigmoid()In [ ]python · cell 29
python
(acts[:,0]-acts[:,1]).sigmoid()In [ ]python · cell 30
python
sm_acts = torch.softmax(acts, dim=1)
sm_actsLog Likelihood
In [ ]python · cell 32
python
targ = tensor([0,1,0,1,1,0])In [ ]python · cell 33
python
sm_actsIn [ ]python · cell 34
python
idx = range(6)
sm_acts[idx, targ]In [ ]python · cell 35
python
from IPython.display import HTML
df = pd.DataFrame(sm_acts, columns=["3","7"])
df['targ'] = targ
df['idx'] = idx
df['result'] = sm_acts[range(6), targ]
t = df.style.hide_index()
#To have html code compatible with our script
html = t._repr_html_().split('</style>')[1]
html = re.sub(r'<table id="([^"]+)"\s*>', r'<table >', html)
display(HTML(html))In [ ]python · cell 36
python
-sm_acts[idx, targ]In [ ]python · cell 37
python
F.nll_loss(sm_acts, targ, reduction='none')Taking the Log
Recall that cross entropy loss may involve the multiplication of many numbers. Multiplying lots of negative numbers together can cause problems like numerical underflow in computers. Therefore, we want to transform these probabilities to larger values so we can perform mathematical operations on them. There is a mathematical function that does exactly this: the logarithm (available as torch.log). It is not defined for numbers less than 0, and looks like this between 0 and 1:
In [ ]python · cell 39
python
plot_function(torch.log, min=0,max=1, ty='log(x)', tx='x')In [ ]python · cell 40
python
plot_function(lambda x: -1*torch.log(x), min=0,max=1, tx='x', ty='- log(x)', title = 'Log Loss when true label = 1')In [ ]python · cell 41
python
from IPython.display import HTML
df['loss'] = -torch.log(tensor(df['result']))
t = df.style.hide_index()
#To have html code compatible with our script
html = t._repr_html_().split('</style>')[1]
html = re.sub(r'<table id="([^"]+)"\s*>', r'<table >', html)
display(HTML(html))Negative Log Likelihood
In [ ]python · cell 43
python
loss_func = nn.CrossEntropyLoss()In [ ]python · cell 44
python
loss_func(acts, targ)In [ ]python · cell 45
python
F.cross_entropy(acts, targ)In [ ]python · cell 46
python
nn.CrossEntropyLoss(reduction='none')(acts, targ)Model Interpretation
In [ ]python · cell 48
python
interp = ClassificationInterpretation.from_learner(learn)
interp.plot_confusion_matrix(figsize=(12,12), dpi=60)In [ ]python · cell 49
python
interp.most_confused(min_val=5)Improving Our Model
The Learning Rate Finder
In [ ]python · cell 52
python
learn = vision_learner(dls, resnet34, metrics=error_rate)
learn.fine_tune(1, base_lr=0.1)In [ ]python · cell 53
python
learn = vision_learner(dls, resnet34, metrics=error_rate)
lr_min,lr_steep = learn.lr_find(suggest_funcs=(minimum, steep))In [ ]python · cell 54
python
print(f"Minimum/10: {lr_min:.2e}, steepest point: {lr_steep:.2e}")In [ ]python · cell 55
python
learn = vision_learner(dls, resnet34, metrics=error_rate)
learn.fine_tune(2, base_lr=3e-3)Unfreezing and Transfer Learning
In [ ]python · cell 57
python
learn.fine_tune??In [ ]python · cell 58
python
learn = vision_learner(dls, resnet34, metrics=error_rate)
learn.fit_one_cycle(3, 3e-3)In [ ]python · cell 59
python
learn.unfreeze()In [ ]python · cell 60
python
learn.lr_find()In [ ]python · cell 61
python
learn.fit_one_cycle(6, lr_max=1e-5)Discriminative Learning Rates
In [ ]python · cell 63
python
learn = vision_learner(dls, resnet34, metrics=error_rate)
learn.fit_one_cycle(3, 3e-3)
learn.unfreeze()
learn.fit_one_cycle(12, lr_max=slice(1e-6,1e-4))In [ ]python · cell 64
python
learn.recorder.plot_loss()Selecting the Number of Epochs
Deeper Architectures
In [ ]python · cell 67
python
from fastai.callback.fp16 import *
learn = vision_learner(dls, resnet50, metrics=error_rate).to_fp16()
learn.fine_tune(6, freeze_epochs=3)Conclusion
Questionnaire
- Why do we first resize to a large size on the CPU, and then to a smaller size on the GPU?
- If you are not familiar with regular expressions, find a regular expression tutorial, and some problem sets, and complete them. Have a look on the book's website for suggestions.
- What are the two ways in which data is most commonly provided, for most deep learning datasets?
- Look up the documentation for
Land try using a few of the new methods that it adds. - Look up the documentation for the Python
pathlibmodule and try using a few methods of thePathclass. - Give two examples of ways that image transformations can degrade the quality of the data.
- What method does fastai provide to view the data in a
DataLoaders? - What method does fastai provide to help you debug a
DataBlock? - Should you hold off on training a model until you have thoroughly cleaned your data?
- What are the two pieces that are combined into cross-entropy loss in PyTorch?
- What are the two properties of activations that softmax ensures? Why is this important?
- When might you want your activations to not have these two properties?
- Calculate the
expandsoftmaxcolumns of <<bear_softmax>> yourself (i.e., in a spreadsheet, with a calculator, or in a notebook). - Why can't we use
torch.whereto create a loss function for datasets where our label can have more than two categories? - What is the value of log(-2)? Why?
- What are two good rules of thumb for picking a learning rate from the learning rate finder?
- What two steps does the
fine_tunemethod do? - In Jupyter Notebook, how do you get the source code for a method or function?
- What are discriminative learning rates?
- How is a Python
sliceobject interpreted when passed as a learning rate to fastai? - Why is early stopping a poor choice when using 1cycle training?
- What is the difference between
resnet50andresnet101? - What does
to_fp16do?
Further Research
- Find the paper by Leslie Smith that introduced the learning rate finder, and read it.
- See if you can improve the accuracy of the classifier in this chapter. What's the best accuracy you can achieve? Look on the forums and the book's website to see what other students have achieved with this dataset, and how they did it.
In [ ]python · cell 73
python