add siameseNet

ConstracstiveLoss.py

@@ -0,0 +1,12 @@
+from torch.nn import functional as F
+import torch
+class ContrastiveLoss(torch.nn.Module):
+
+    def __init__(self, margin=2.0):
+        super(ContrastiveLoss, self).__init__()
+        self.margin=margin
+    def forward(self, output1,output2,label):
+        euclidean_distance = F.pairwise_distance(output1, output2)
+        loss_contrastive = torch.mean((1-label) * torch.pow(euclidean_distance, 2) +
+                                      (label) * torch.pow(torch.clamp(self.margin - euclidean_distance, min=0.0), 2))
+        return loss_contrastive

SiameseNetworkDataset.py

@@ -0,0 +1,26 @@
+from torch.utils.data import Dataset
+import torch
+import numpy as np
+from PIL import Image
+class SiameseNetworkDataset(Dataset):
+    def __init__(self, root_dir, label, transform=None):
+        self.root_dir = root_dir
+        self.label = label
+        self.transform=transform
+    def __getitem__(self, index):
+
+        img0 = Image.open(self.root_dir + '/pair1/' + str(index) + '.bmp')
+        img1 = Image.open(self.root_dir + '/pair2/' + str(index) + '.bmp')
+
+        img0=img0.convert("L")
+        img1=img1.convert("L")
+
+        if self.transform is not None:
+            img0=self.transform(img0)
+            img1=self.transform(img1)
+
+        return img0,img1, torch.from_numpy(np.array([int(self.label[index])],dtype=np.float32))
+    def __len__(self):
+        return len(self.label)
+
+

compute_desc.py

@@ -0,0 +1,19 @@
+from net1 import SiameseNetwork
+from torchvision import transforms
+from PIL import Image
+import torch
+from torch.autograd import Variable
+
+net=SiameseNetwork()
+
+net.load_state_dict(torch.load('model.pt'))
+transform=transforms.ToTensor()
+img=Image.open('tmp.bmp')
+img=img.convert('L')
+img=transform(img)
+img=img.unsqueeze(0)
+img.view(img.size()[0],-1)
+_,output=net(Variable(img),Variable(img))
+print output
+
+

extract_data.py

@@ -0,0 +1,48 @@
+
+from scipy import misc
+
+imglist='liberty/m50_200000_200000_0.txt'
+
+with open(imglist,'r') as f:
+    content=f.read();
+
+lines=content.split('\n')
+
+num=len(lines)-1
+
+f.close()
+
+f=open('label.txt','w');
+
+for i in range(num):
+    if (i%1000==0):
+       print i
+
+    list=lines[i].split(' ')
+    id1=int(list[0])
+    id2=int(list[3])
+    label=(list[1]==list[4])
+
+
+    #first patch
+    m=id1/256
+    count=(id1)%256
+    x=(count)%16
+    y=count/16
+    img=misc.imread('liberty/patches'+str(m).zfill(4)+'.bmp')
+    x_=x*64
+    y_=y*64
+    patch = img[y_:y_ + 64, x_:x_ + 64]
+    misc.imsave('pair1/'+str(i)+'.bmp',patch)
+
+    #second patch
+    m=id2/256
+    count=(id2)%256
+    x=(count)%16
+    y=count/16
+    img=misc.imread('liberty/patches'+str(m).zfill(4)+'.bmp')
+    x_=x*64
+    y_=y*64
+    patch=img[y_:y_+64,x_:x_+64]
+    misc.imsave('pair2/'+str(i)+'.bmp',patch)
+    f.write(str(int(label)))

net1.py

@@ -0,0 +1,30 @@
+import torch.nn as nn
+
+class SiameseNetwork(nn.Module):
+    def __init__(self):
+        super(SiameseNetwork, self).__init__()
+        self.cnn1 = nn.Sequential(
+            nn.Conv2d(1,32, kernel_size=7),
+            nn.MaxPool2d(2,stride=2),
+            nn.ReLU(inplace=True),
+            nn.Conv2d(32, 64, kernel_size= 6),
+            nn.MaxPool2d(3,stride=3),
+            nn.ReLU(inplace=True),
+            nn.Conv2d(64,128,kernel_size= 5,),
+            nn.MaxPool2d(4, stride=4)
+        )
+        self.fc1=nn.Sequential(
+            nn.Linear(128,128)
+        )
+    def forward_once(self,x):
+        output=self.cnn1(x)
+        output=output.view(output.size()[0],-1)
+        output =self.fc1(output)
+        return output
+
+    def forward(self, input1,input2):
+        output1=self.forward_once(input1)
+        output2=self.forward_once(input2)
+        return output1,output2
+
+

visualize_data.py

@@ -0,0 +1,30 @@
+from torch.utils.data import DataLoader
+import torchvision
+import torch
+import  matplotlib.pyplot as plt
+import torchvision.transforms as transforms
+import numpy as np
+import torchvision.datasets as dset
+from SiameseNetworkDataset import SiameseNetworkDataset
+
+
+
+def imshow(img,text,should_save=False):
+    npimg = img.numpy()
+    plt.axis("off")
+    if text:
+        plt.text(75, 8, text, style='italic',fontweight='bold',
+            bbox={'facecolor':'white', 'alpha':0.8, 'pad':10})
+    plt.imshow(np.transpose(npimg, (1, 2, 0)))
+    plt.show()
+
+f=open('label.txt','r');
+label=f.read()
+dataset=SiameseNetworkDataset('.',label,transform=transforms.ToTensor())
+
+vis_dataloader=DataLoader(dataset,shuffle=True,batch_size=8)
+dataiter=iter(vis_dataloader)
+example_batch = next(dataiter)
+concatenated = torch.cat((example_batch[0],example_batch[1]),0)
+imshow(torchvision.utils.make_grid(concatenated),'img')
+print(example_batch[2].numpy())