Human Protine Multi Label Image Classification
In [1]:
import os
import torch
import pandas as pd
import numpy as np
from torch.utils.data import Dataset , random_split,DataLoader
from PIL import Image
import torchvision.models as models
import matplotlib.pyplot as plt
import torchvision.transforms as transforms
from sklearn.metrics import f1_score
import torch.nn.functional as F
import torch.nn as nn
from torchvision.utils import make_grid
%matplotlib inline
Dara is already included in ../input folder
In [2]:
Data_dir = '../input/jovian-pytorch-z2g/Human protein atlas/'
Train_dir = Data_dir + '/train' # training images
Test_dir = Data_dir + '/test' # test images
Train_csv = Data_dir + '/train.csv' # real labels for training images
Test_csv = '../input/jovian-pytorch-z2g/submission.csv' #dummy labels for test images
This is a multi-label classification problem. Each image can have more than one type of protein.
In [3]:
!head "{Train_csv}"
we don't have true labels for the test set. Privided labels are only dummmy labels . We need to replace those dummy labels with the labels which our models will predict
In [4]:
!head "{Test_csv}"
images are named as imageid.png
In [5]:
!ls "{Train_dir}" | head
In [6]:
train_df = pd.read_csv(Train_csv)
In [7]:
train_df.head()
Out[7]:
In [8]:
train_df.count() #no of images
Out[8]:
In [9]:
labels = {0: 'Mitochondria',
1: 'Nuclear bodies',
2: 'Nucleoli',
3: 'Golgi apparatus',
4: 'Nucleoplasm',
5: 'Nucleoli fibrillar center',
6: 'Cytosol',
7: 'Plasma membrane',
8: 'Centrosome',
9: 'Nuclear speckles'}
encoding the labels: creating tensors from labels as vector of 1 and 0.
creating 10 zeros and putting 1 at the label number positions.
In [10]:
def encode_label(label):
target = torch.zeros(10)
for l in str(label).split(' '):
target[int(l)] = 1
return target
In [11]:
def decode_target(target,text_labels=False,threshold=0.5):
result = []
for i,x in enumerate(target):
if (x>threshold):
if text_labels:
result.append(labels[i]+"("+str(i)+")")
else:
result.append(str(i))
return ' '.join(result)
In [12]:
encode_label('2 4 5')
Out[12]:
In [13]:
decode_target(torch.tensor([0, 0, 1, 0, 1, 1, 0, 0, 0, 0.]))
Out[13]:
In [14]:
decode_target(torch.tensor([0, 0, 1, 0, 1, 1, 0, 0, 0, 0.]), text_labels=True)
Out[14]:
Dataset and Dataloaders¶
In [15]:
class HumanProteinDataset(Dataset):
def __init__(self, csv_file, root_dir, transform=None):
self.df = pd.read_csv(csv_file)
self.transform = transform
self.root_dir = root_dir
def __len__(self):
return len(self.df)
def __getitem__(self, idx):
row = self.df.loc[idx]
img_id, img_label = row['Image'], row['Label']
img_fname = self.root_dir + "/" + str(img_id) + ".png"
img = Image.open(img_fname)
if self.transform:
img = self.transform(img)
return img, encode_label(img_label)
transforms.Resize(128) before transforms.ToTensor() to resize images to size 128x128 before converting them into tensors.
In [16]:
transform = transforms.Compose([transforms.ToTensor()])
dataset = HumanProteinDataset(Train_csv,Train_dir,transform=transform)
In [17]:
len(dataset)
Out[17]:
plotting images
In [18]:
def show_sample(img, target, invert=True):
if invert:
plt.imshow(1 - img.permute((1, 2, 0)))
else:
plt.imshow(img.permute(1, 2, 0))
print('Labels:', decode_target(target, text_labels=True))
In [19]:
show_sample(*dataset[3],invert=False)
train test split
In [20]:
torch.manual_seed(10)
Out[20]:
In [21]:
val_pct = 0.2
val_size = int(val_pct * len(dataset))
train_size = len(dataset) - val_size
In [22]:
train_ds, val_ds = random_split(dataset, [train_size, val_size])
len(train_ds), len(val_ds)
Out[22]:
In [23]:
batch_size = 50
In [24]:
train_dl = DataLoader(train_ds, batch_size, shuffle=True, num_workers=2, pin_memory=True)
val_dl = DataLoader(val_ds, batch_size*2, num_workers=2, pin_memory=True)
In [25]:
def show_batch(dl, invert=True):
for images, labels in dl:
fig, ax = plt.subplots(figsize=(10, 8))
ax.set_xticks([]); ax.set_yticks([])
data = 1-images if invert else images
ax.imshow(make_grid(data, nrow=10).permute(1, 2, 0))
break
In [26]:
show_batch(train_dl)
model¶
In [27]:
def F_score(output, label, threshold=0.5, beta=1):
prob = output > threshold
label = label > threshold
TP = (prob & label).sum(1).float()
TN = ((~prob) & (~label)).sum(1).float()
FP = (prob & (~label)).sum(1).float()
FN = ((~prob) & label).sum(1).float()
precision = torch.mean(TP / (TP + FP + 1e-12))
recall = torch.mean(TP / (TP + FN + 1e-12))
F2 = (1 + beta**2) * precision * recall / (beta**2 * precision + recall + 1e-12)
return F2.mean(0)
In [28]:
class MultilabelImageClassificationBase(nn.Module):
def training_step(self, batch):
images, targets = batch
out = self(images)
loss = F.binary_cross_entropy(out, targets)
return loss
def validation_step(self, batch):
images, targets = batch
out = self(images) # Generate predictions
loss = F.binary_cross_entropy(out, targets) # Calculate loss
score = F_score(out, targets)
return {'val_loss': loss.detach(), 'val_score': score.detach() }
def validation_epoch_end(self, outputs):
batch_losses = [x['val_loss'] for x in outputs]
epoch_loss = torch.stack(batch_losses).mean() # Combine losses
batch_scores = [x['val_score'] for x in outputs]
epoch_score = torch.stack(batch_scores).mean() # Combine accuracies
return {'val_loss': epoch_loss.item(), 'val_score': epoch_score.item()}
def epoch_end(self, epoch, result):
print("Epoch [{}], train_loss: {:.4f}, val_loss: {:.4f}, val_score: {:.4f}".format(
epoch, result['train_loss'], result['val_loss'], result['val_score']))
In [29]:
class ProteinCnnModel(MultilabelImageClassificationBase):
def __init__(self):
super().__init__()
self.network = nn.Sequential(
nn.Conv2d(3, 32, kernel_size=3, padding=1),
nn.ReLU(),
nn.MaxPool2d(2, 2),
nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1),
nn.ReLU(),
nn.MaxPool2d(2, 2),
nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1),
nn.ReLU(),
nn.MaxPool2d(2, 2),
nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1),
nn.ReLU(),
nn.MaxPool2d(2, 2),
nn.Conv2d(128, 128, kernel_size=3, stride=1, padding=1),
nn.ReLU(),
nn.MaxPool2d(2, 2),
nn.Conv2d(128, 256, kernel_size=3, stride=1, padding=1),
nn.ReLU(),
nn.AdaptiveAvgPool2d(1),
nn.Flatten(),
nn.Linear(256, 128),
nn.ReLU(),
nn.Linear(128, 64),
nn.ReLU(),
nn.Linear(64, 10),
nn.Sigmoid()
)
def forward(self, xb):
return self.network(xb)
In [30]:
class ProteinCnnModel2(MultilabelImageClassificationBase):
def __init__(self):
super().__init__()
# Use a pretrained model
self.network = models.resnet34(pretrained=True)
# Replace last layer
num_ftrs = self.network.fc.in_features
self.network.fc = nn.Linear(num_ftrs, 10)
def forward(self, xb):
return torch.sigmoid(self.network(xb))
In [31]:
model = ProteinCnnModel2()
model
Out[31]:
In [32]:
def get_default_device():
"""Pick GPU if available, else CPU"""
if torch.cuda.is_available():
return torch.device('cuda')
else:
return torch.device('cpu')
def to_device(data, device):
"""Move tensor(s) to chosen device"""
if isinstance(data, (list,tuple)):
return [to_device(x, device) for x in data]
return data.to(device, non_blocking=True)
class DeviceDataLoader():
"""Wrap a dataloader to move data to a device"""
def __init__(self, dl, device):
self.dl = dl
self.device = device
def __iter__(self):
"""Yield a batch of data after moving it to device"""
for b in self.dl:
yield to_device(b, self.device)
def __len__(self):
"""Number of batches"""
return len(self.dl)
In [33]:
device = get_default_device()
device
Out[33]:
In [34]:
train_dl = DeviceDataLoader(train_dl, device)
val_dl = DeviceDataLoader(val_dl, device)
to_device(model, device);
In [35]:
def try_batch(dl):
for images, labels in dl:
print('images.shape:', images.shape)
out = model(images)
print('out.shape:', out.shape)
print('out[0]:', out[0])
break
try_batch(train_dl)
trin the model¶
In [36]:
from tqdm.notebook import tqdm
In [37]:
@torch.no_grad()
def evaluate(model, val_loader):
model.eval()
outputs = [model.validation_step(batch) for batch in val_loader]
return model.validation_epoch_end(outputs)
def fit(epochs, lr, model, train_loader, val_loader, opt_func=torch.optim.SGD):
torch.cuda.empty_cache()
history = []
optimizer = opt_func(model.parameters(), lr)
for epoch in range(epochs):
# Training Phase
model.train()
train_losses = []
for batch in tqdm(train_loader):
loss = model.training_step(batch)
train_losses.append(loss)
loss.backward()
optimizer.step()
optimizer.zero_grad()
# Validation phase
result = evaluate(model, val_loader)
result['train_loss'] = torch.stack(train_losses).mean().item()
model.epoch_end(epoch, result)
history.append(result)
return history
In [38]:
model = to_device(ProteinCnnModel2(), device)
In [39]:
evaluate(model, val_dl)
Out[39]:
In [43]:
num_epochs = 5
opt_func = torch.optim.Adam
lr = 1e-5
In [44]:
history = fit(num_epochs, lr, model, train_dl, val_dl, opt_func)
In [1]:
!pip install jovian
In [3]:
import jovian
In [5]:
jovian.commit()
Out[5]:
In [ ]: