Add files via upload

Author: abdullahbas

PyTorch/PyTorchAugmentation.py

@@ -0,0 +1,118 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Sat Feb 13 19:24:17 2021
+
+@author: trabz
+"""
+import torch 
+
+
+from xmlParser import Parser
+import glob
+import numpy as np
+from matplotlib import pyplot as plt
+import cv2
+import torchvision.transforms as transforms
+class dataloader():
+    
+    def __init__(self,path,transform=None):
+        
+        self.paths=glob.glob(path)
+        self.transform=transform
+        
+        
+    def __len__(self):
+        return len(self.paths)
+    
+    def __getitem__(self,idx):
+        
+        annotation=(Parser.myType(self.paths[idx],idx,classes=['bird','zebra']))
+        image=plt.imread(self.paths[annotation['image_id']].replace('xml','jpg'))
+        
+        if self.transform is not None:
+            
+            image_aug=self.transform(image)
+        
+        return image,image_aug,annotation
+
+
+def collate_fn(batch):
+            return tuple(zip(*batch))
+
+
+data_transform=transforms.Compose([
+    transforms.ToTensor(),
+    #transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
+    transforms.ToPILImage(),
+    transforms.RandomApply([transforms.CenterCrop(222), transforms.RandomCrop(170)],p=0.01),
+   
+    
+    transforms.RandomAffine((-15,15)),
+        
+
+    
+    #transforms.Grayscale(),
+    transforms.RandomGrayscale(p=0.1),
+    transforms.RandomHorizontalFlip(p=0.1),
+    transforms.RandomVerticalFlip(p=0.1),
+    transforms.RandomApply([transforms.ColorJitter(brightness=15,contrast=12,hue=0.2)],p=0.1),
+    # transforms.RandomResizedCrop((200,200)),
+    transforms.RandomRotation((-90,90)),
+    transforms.RandomPerspective(p=0.1),
+    transforms.ToTensor(),
+    # transforms.RandomCrop(10),
+    transforms.RandomErasing(p=0.1),
+
+    
+    # 
+    ])
+
+
+
+
+
+
+path=path= 'Image A*/train/*.xml'
+
+dataset = dataloader(path,data_transform)
+data_loader = torch.utils.data.DataLoader(
+    dataset, batch_size=1, collate_fn=collate_fn)
+
+
+
+
+plt.figure(figsize=(12,12))
+# for idx,(image,imgO,result) in enumerate(data_loader):
+    
+#     #plt.figure()
+#     #plt.imshow(result[0].permute(1,2,0).numpy(),cmap='gray')
+#     img=imgO[0].clone().permute(1,2,0).numpy()
+#     image=image[0]
+#     sz,wd,_=np.array(image.shape)-np.array((img).shape)
+#     img2=img
+#     img=np.pad((img),((sz//2,sz//2),(wd//2,wd//2),(0,0)))
+    
+#     try:
+#         plt.imsave(f'PyTorch/images/{idx}.png',np.hstack(((img*255).astype('uint8'),image)))
+#     except ValueError:
+#         img=np.dstack((img,img,img))
+#         plt.imsave(f'PyTorch/images/{idx}.png',np.hstack(((img*255).astype('uint8'),image)))
+#     plt.axis('off')
+#     plt.tight_layout()
+#     #plt.imshow(np.dstack((img[:,:,2],img[:,:,1],img[:,:,0])))
+#     plt.imshow(image)    
+    
+#♣plt.savefig('TFOrig.png',dpi=110,transparent=True)
+
+for idx,(image,imgO,result) in enumerate(data_loader):
+    
+    #plt.figure()
+    #plt.imshow(result[0].permute(1,2,0).numpy(),cmap='gray')
+    plt.subplot(6,5,idx+1)
+    
+    plt.axis('off')
+    plt.tight_layout()
+    #plt.imshow(np.dstack((img[:,:,2],img[:,:,1],img[:,:,0])))
+    plt.imshow(imgO[0].permute(1,2,0).numpy())    
+    
+plt.savefig('PyTorchAugmented2.png',dpi=110,transparent=True)