In [ ]:

%matplotlib inline

Lesson 6: pets revisited¶

In [ ]:

from fastai2.vision.all import *
from nbdev.showdoc import *

In [ ]:

bs = 64

In [ ]:

path = untar_data(URLs.PETS)

Data augmentation¶

In [ ]:

tfms = partial(aug_transforms, max_rotate=20, max_zoom=1.3, max_lighting=0.4, max_warp=0.4,
                      p_affine=1., p_lighting=1.)

In [ ]:

doc(aug_transforms)

In [ ]:

def repeat_one(source, n=128): return [get_image_files(source)[0]]*n

In [ ]:

def get_dls(size, bs, pad_mode='reflection', batch=False):
    pets = DataBlock(blocks=(ImageBlock, CategoryBlock),
                 get_items=repeat_one,
                 splitter=RandomSplitter(0.2, seed=2),
                 get_y=RegexLabeller(r'([^/]+)_\d+.jpg$'),
                 item_tfms=Resize(460),
                 batch_tfms=[*tfms(size=size, pad_mode=pad_mode, batch=batch), Normalize.from_stats(*imagenet_stats)])
    return pets.dataloaders(path/'images', path=path, bs=bs)

In [ ]:

dls = get_dls(224, bs, 'zeros')

In [ ]:

dls.show_batch(max_n=9, figsize=(8,8))

In [ ]:

dls = get_dls(224,bs)

In [ ]:

dls.show_batch(max_n=9, figsize=(8,8))

batch=True means we pick the same random augmentation for all the images in the batch.

In [ ]:

dls = get_dls(224,bs,batch=True)

In [ ]:

dls.show_batch(max_n=9, figsize=(8,8))

Train a model¶

In [ ]:

def get_dls(size, bs, pad_mode='reflection'):
    pets = DataBlock(blocks=(ImageBlock, CategoryBlock),
                 get_items=get_image_files,
                 splitter=RandomSplitter(0.2, seed=2),
                 get_y=RegexLabeller(r'([^/]+)_\d+.jpg$'),
                 item_tfms=RandomResizedCrop(460, min_scale=0.75),
                 batch_tfms=[*tfms(size=size, pad_mode=pad_mode), Normalize.from_stats(*imagenet_stats)])
    return pets.dataloaders(path/'images', path=path, bs=bs)

In [ ]:

dls = get_dls(224,bs)

In [ ]:

learn = cnn_learner(dls, resnet34, metrics=error_rate, config=cnn_config(bn_final=True))

In [ ]:

learn.fit_one_cycle(3, slice(1e-2), pct_start=0.8)

epoch	train_loss	valid_loss	error_rate	time
0	1.680457	0.346096	0.080514	00:20
1	0.994624	0.260845	0.071719	00:15
2	0.730085	0.237081	0.069012	00:15

In [ ]:

learn.unfreeze()
learn.fit_one_cycle(2, lr_max=slice(1e-6,1e-3), pct_start=0.8)

epoch	train_loss	valid_loss	error_rate	time
0	0.596662	0.227051	0.058863	00:19
1	0.575870	0.229266	0.059540	00:18

In [ ]:

dls = get_dls(352,bs)
learn.dls = dls

In [ ]:

learn.fit_one_cycle(2, lr_max=slice(1e-6,1e-4))

epoch	train_loss	valid_loss	error_rate	time
0	0.513675	0.213754	0.051421	00:35
1	0.452948	0.209984	0.051421	00:34

In [ ]:

learn.save('352')

Convolution kernel¶

In [ ]:

dls = get_dls(352,16)

In [ ]:

learn = cnn_learner(dls, resnet34, metrics=error_rate, config=cnn_config(bn_final=True)).load('352')

In [ ]:

idx=0
x,y = dls.valid_ds[idx]
show_at(dls.valid_ds, idx);

In [ ]:

k = tensor([
    [0.  ,-5/3,1],
    [-5/3,-5/3,1],
    [1.  ,1   ,1],
]).expand(1,3,3,3)/6

In [ ]:

Out[ ]:

tensor([[[[ 0.0000, -0.2778,  0.1667],
          [-0.2778, -0.2778,  0.1667],
          [ 0.1667,  0.1667,  0.1667]],

         [[ 0.0000, -0.2778,  0.1667],
          [-0.2778, -0.2778,  0.1667],
          [ 0.1667,  0.1667,  0.1667]],

         [[ 0.0000, -0.2778,  0.1667],
          [-0.2778, -0.2778,  0.1667],
          [ 0.1667,  0.1667,  0.1667]]]])

In [ ]:

k.shape

Out[ ]:

torch.Size([1, 3, 3, 3])

In [ ]:

t = tensor(x).permute(2,0,1).float(); t.shape

Out[ ]:

torch.Size([3, 233, 300])

In [ ]:

t[None].shape

Out[ ]:

torch.Size([1, 3, 233, 300])

In [ ]:

edge = F.conv2d(t[None], k)

In [ ]:

show_image(edge[0], figsize=(5,5));

In [ ]:

dls.c

Out[ ]:

In [ ]:

learn.model

Out[ ]:

Sequential(
  (0): Sequential(
    (0): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU(inplace=True)
    (3): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
    (4): Sequential(
      (0): BasicBlock(
        (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (1): BasicBlock(
        (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (2): BasicBlock(
        (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (5): Sequential(
      (0): BasicBlock(
        (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (downsample): Sequential(
          (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)
          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BasicBlock(
        (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (2): BasicBlock(
        (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (3): BasicBlock(
        (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (6): Sequential(
      (0): BasicBlock(
        (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (downsample): Sequential(
          (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)
          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BasicBlock(
        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (2): BasicBlock(
        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (3): BasicBlock(
        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (4): BasicBlock(
        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (5): BasicBlock(
        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (7): Sequential(
      (0): BasicBlock(
        (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (downsample): Sequential(
          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BasicBlock(
        (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
      (2): BasicBlock(
        (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
  )
  (1): Sequential(
    (0): AdaptiveConcatPool2d(
      (ap): AdaptiveAvgPool2d(output_size=1)
      (mp): AdaptiveMaxPool2d(output_size=1)
    )
    (1): Flatten()
    (2): BatchNorm1d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (3): Dropout(p=0.25, inplace=False)
    (4): Linear(in_features=1024, out_features=512, bias=False)
    (5): ReLU(inplace=True)
    (6): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (7): Dropout(p=0.5, inplace=False)
    (8): Linear(in_features=512, out_features=37, bias=False)
    (9): BatchNorm1d(37, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
  )
)

In [ ]:

print(learn.summary())

Sequential (Input shape: 16 x 3 x 352 x 352)
================================================================
Layer (type)         Output Shape         Param #    Trainable 
================================================================
Conv2d               16 x 64 x 176 x 176  9,408      False     
________________________________________________________________
BatchNorm2d          16 x 64 x 176 x 176  128        True      
________________________________________________________________
ReLU                 16 x 64 x 176 x 176  0          False     
________________________________________________________________
MaxPool2d            16 x 64 x 88 x 88    0          False     
________________________________________________________________
Conv2d               16 x 64 x 88 x 88    36,864     False     
________________________________________________________________
BatchNorm2d          16 x 64 x 88 x 88    128        True      
________________________________________________________________
ReLU                 16 x 64 x 88 x 88    0          False     
________________________________________________________________
Conv2d               16 x 64 x 88 x 88    36,864     False     
________________________________________________________________
BatchNorm2d          16 x 64 x 88 x 88    128        True      
________________________________________________________________
Conv2d               16 x 64 x 88 x 88    36,864     False     
________________________________________________________________
BatchNorm2d          16 x 64 x 88 x 88    128        True      
________________________________________________________________
ReLU                 16 x 64 x 88 x 88    0          False     
________________________________________________________________
Conv2d               16 x 64 x 88 x 88    36,864     False     
________________________________________________________________
BatchNorm2d          16 x 64 x 88 x 88    128        True      
________________________________________________________________
Conv2d               16 x 64 x 88 x 88    36,864     False     
________________________________________________________________
BatchNorm2d          16 x 64 x 88 x 88    128        True      
________________________________________________________________
ReLU                 16 x 64 x 88 x 88    0          False     
________________________________________________________________
Conv2d               16 x 64 x 88 x 88    36,864     False     
________________________________________________________________
BatchNorm2d          16 x 64 x 88 x 88    128        True      
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   73,728     False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
ReLU                 16 x 128 x 44 x 44   0          False     
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   147,456    False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   8,192      False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   147,456    False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
ReLU                 16 x 128 x 44 x 44   0          False     
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   147,456    False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   147,456    False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
ReLU                 16 x 128 x 44 x 44   0          False     
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   147,456    False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   147,456    False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
ReLU                 16 x 128 x 44 x 44   0          False     
________________________________________________________________
Conv2d               16 x 128 x 44 x 44   147,456    False     
________________________________________________________________
BatchNorm2d          16 x 128 x 44 x 44   256        True      
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   294,912    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
ReLU                 16 x 256 x 22 x 22   0          False     
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   32,768     False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
ReLU                 16 x 256 x 22 x 22   0          False     
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
ReLU                 16 x 256 x 22 x 22   0          False     
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
ReLU                 16 x 256 x 22 x 22   0          False     
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
ReLU                 16 x 256 x 22 x 22   0          False     
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
ReLU                 16 x 256 x 22 x 22   0          False     
________________________________________________________________
Conv2d               16 x 256 x 22 x 22   589,824    False     
________________________________________________________________
BatchNorm2d          16 x 256 x 22 x 22   512        True      
________________________________________________________________
Conv2d               16 x 512 x 11 x 11   1,179,648  False     
________________________________________________________________
BatchNorm2d          16 x 512 x 11 x 11   1,024      True      
________________________________________________________________
ReLU                 16 x 512 x 11 x 11   0          False     
________________________________________________________________
Conv2d               16 x 512 x 11 x 11   2,359,296  False     
________________________________________________________________
BatchNorm2d          16 x 512 x 11 x 11   1,024      True      
________________________________________________________________
Conv2d               16 x 512 x 11 x 11   131,072    False     
________________________________________________________________
BatchNorm2d          16 x 512 x 11 x 11   1,024      True      
________________________________________________________________
Conv2d               16 x 512 x 11 x 11   2,359,296  False     
________________________________________________________________
BatchNorm2d          16 x 512 x 11 x 11   1,024      True      
________________________________________________________________
ReLU                 16 x 512 x 11 x 11   0          False     
________________________________________________________________
Conv2d               16 x 512 x 11 x 11   2,359,296  False     
________________________________________________________________
BatchNorm2d          16 x 512 x 11 x 11   1,024      True      
________________________________________________________________
Conv2d               16 x 512 x 11 x 11   2,359,296  False     
________________________________________________________________
BatchNorm2d          16 x 512 x 11 x 11   1,024      True      
________________________________________________________________
ReLU                 16 x 512 x 11 x 11   0          False     
________________________________________________________________
Conv2d               16 x 512 x 11 x 11   2,359,296  False     
________________________________________________________________
BatchNorm2d          16 x 512 x 11 x 11   1,024      True      
________________________________________________________________
AdaptiveAvgPool2d    16 x 512 x 1 x 1     0          False     
________________________________________________________________
AdaptiveMaxPool2d    16 x 512 x 1 x 1     0          False     
________________________________________________________________
Flatten              16 x 1024            0          False     
________________________________________________________________
BatchNorm1d          16 x 1024            2,048      True      
________________________________________________________________
Dropout              16 x 1024            0          False     
________________________________________________________________
Linear               16 x 512             524,288    True      
________________________________________________________________
ReLU                 16 x 512             0          False     
________________________________________________________________
BatchNorm1d          16 x 512             1,024      True      
________________________________________________________________
Dropout              16 x 512             0          False     
________________________________________________________________
Linear               16 x 37              18,944     True      
________________________________________________________________
BatchNorm1d          16 x 37              74         True      
________________________________________________________________

Total params: 21,831,050
Total trainable params: 563,402
Total non-trainable params: 21,267,648

Optimizer used: <function Adam at 0x7f7778f8b440>
Loss function: FlattenedLoss of CrossEntropyLoss()

Model frozen up to parameter group number 2

Callbacks:
  - TrainEvalCallback
  - Recorder
  - ProgressCallback

Heatmap¶

In [ ]:

m = learn.model.eval();

In [ ]:

b = dls.one_batch()
xb_im = TensorImage(dls.train.decode(b)[0][0])
xb = b[0]

In [ ]:

def hooked_backward(cat=y):
    with hook_output(m[0]) as hook_a: 
        with hook_output(m[0], grad=True) as hook_g:
            preds = m(xb)
            preds[0,int(cat)].backward()
    return hook_a,hook_g

In [ ]:

hook_a,hook_g = hooked_backward()

In [ ]:

acts  = hook_a.stored[0].cpu()
acts.shape

Out[ ]:

torch.Size([512, 11, 11])

In [ ]:

avg_acts = acts.mean(0)
avg_acts.shape

Out[ ]:

torch.Size([11, 11])

In [ ]:

def show_heatmap(hm):
    _,ax = plt.subplots()
    xb_im.show(ctx=ax)
    ax.imshow(hm, alpha=0.6, extent=(0,352,352,0),
              interpolation='bilinear', cmap='magma');

In [ ]:

show_heatmap(avg_acts)

Grad-CAM¶

Paper: Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization

In [ ]:

grad = hook_g.stored[0][0].cpu()
grad_chan = grad.mean(1).mean(1)
grad.shape,grad_chan.shape

Out[ ]:

(torch.Size([512, 11, 11]), torch.Size([512]))

In [ ]:

mult = (acts*grad_chan[...,None,None]).mean(0)

In [ ]:

show_heatmap(mult)

In [ ]:

fn = Path.home()/'tmp/bulldog_maine.png' #Replace with your own image

In [ ]:

x = PILImage.create(fn); x

Out[ ]:

In [ ]:

dl = dls.test_dl([fn])
b = dl.one_batch()
xb_im = TensorImage(dls.train.decode(b)[0][0])
xb = b[0]

In [ ]:

hook_a,hook_g = hooked_backward()

In [ ]:

acts = hook_a.stored[0].cpu()
grad = hook_g.stored[0][0].cpu()

grad_chan = grad.mean(1).mean(1)
mult = (acts*grad_chan[...,None,None]).mean(0)

In [ ]:

show_heatmap(mult)

In [ ]:

dls.vocab[0]

Out[ ]:

'Abyssinian'

In [ ]:

hook_a,hook_g = hooked_backward(0)

In [ ]:

acts = hook_a.stored[0].cpu()
grad = hook_g.stored[0][0].cpu()

grad_chan = grad.mean(1).mean(1)
mult = (acts*grad_chan[...,None,None]).mean(0)

In [ ]:

show_heatmap(mult)

fin¶

In [ ]: