vm get augmented data

@@ -10,7 +10,7 @@ from torch.utils.tensorboard import SummaryWriter
 from utils import *
 
 # command
-# python "eval.py" --model_path='logs/'
+# python eval.py --model_path='logs/April_16_00:26:10__resnet50__None/'
 
 def eval(model_path):
     print('\n[+] Parse arguments')
@@ -34,9 +34,17 @@ def eval(model_path):
 
     print('\n[+] Load dataset')
     test_transform = get_valid_transform(args, model)
-    test_dataset = get_dataset(args, test_transform, 'test')
-    #print("len(dataset): ",  len(test_dataset), type(test_dataset)) # 590
-    
+    #print('\nTEST Transform\n', test_transform)
+    test_dataset = get_dataset(args, test_transform, 'test') 
+
+    """
+    test_transform
+    Compose(
+        Resize(size=[224, 224], interpolation=PIL.Image.BILINEAR)
+        ToTensor()
+    )
+    """
+   
     test_loader = iter(get_dataloader(args, test_dataset)) ###
 
     print('\n[+] Start testing')

@@ -17,15 +17,15 @@ from utils import *
 DEFALUT_CANDIDATES = [
     ShearXY,
     TranslateXY,
-    # Rotate,
+    Rotate,
     # AutoContrast,
     # Invert,
-    Equalize,
-    Solarize,
+    Equalize, # Histogram Equalize --> white tumor
+    #Solarize,
     Posterize,
-    Contrast,
+    # Contrast,
     # Color,
-    Brightness,
+    # Brightness,
     Sharpness,
     Cutout,
 #     SamplePairing,
@@ -154,8 +154,9 @@ def search_subpolicies_hyperopt(args, transform_candidates, child_model, dataset
         subpolicy = transforms.Compose([
             ## baseline augmentation
             transforms.Pad(4),
-            transforms.RandomCrop(32),
+           # transforms.RandomCrop(240), #32 ->240
             transforms.RandomHorizontalFlip(),
+            transforms.Resize([240, 240]),
             ## policy
             *subpolicy,
             ## to tensor
@@ -191,8 +192,8 @@ def process_fn(args_str, model, dataset, Dm_indx, Da_indx, T, transform_candidat
 
     return _transform
 
-#fast_auto_augment(args, model, K=4, B=1, num_process=4)
-def fast_auto_augment(args, model, transform_candidates=None, K=5, B=100, T=2, N=10, num_process=5):
+#fast_auto_augment(args, model, K=4, B=100, num_process=4)
+def fast_auto_augment(args, model, transform_candidates=None, K=5, B=100, T=2, N=2, num_process=5):
     args_str = json.dumps(args._asdict())
     dataset = get_dataset(args, None, 'trainval')
     num_process = min(torch.cuda.device_count(), num_process)
@@ -215,6 +216,6 @@ def fast_auto_augment(args, model, transform_candidates=None, K=5, B=100, T=2, N
         for future in futures:
             transform.extend(future.result())
 
-    transform = transforms.RandomChoice(transform)
+    #transform = transforms.RandomChoice(transform)
 
     return transform

+import os
+import fire
+import json
+from pprint import pprint
+import pickle
+
+import torch
+import torch.nn as nn
+from torch.utils.tensorboard import SummaryWriter
+
+from utils import *
+
+# command
+# python getAugmented.py --model_path='logs/April_16_21:50:17__resnet50__None/'
+
+def eval(model_path):
+    print('\n[+] Parse arguments')
+    kwargs_path = os.path.join(model_path, 'kwargs.json')
+    kwargs = json.loads(open(kwargs_path).read())
+    args, kwargs = parse_args(kwargs)
+    pprint(args)
+    device = torch.device('cuda' if args.use_cuda else 'cpu')
+
+    cp_path = os.path.join(model_path, 'augmentation.cp')
+    
+    writer = SummaryWriter(log_dir=model_path)
+
+
+    print('\n[+] Load transform')
+    # list
+    with open(cp_path, 'rb') as f:
+        aug_transform_list = pickle.load(f)
+
+    augmented_image_list = [torch.Tensor(240,0)] * len(get_dataset(args, None, 'test'))
+
+
+    print('\n[+] Load dataset')
+    for aug_idx, aug_transform in enumerate(aug_transform_list):
+        dataset = get_dataset(args, aug_transform, 'test') 
+
+        loader = iter(get_aug_dataloader(args, dataset))
+
+        for i, (images, target) in enumerate(loader):
+            images = images.view(240, 240)
+
+            # concat image
+            augmented_image_list[i] = torch.cat([augmented_image_list[i], images], dim = 1)
+            
+            if i % 1000 == 0: 
+                print("\n images size: ", augmented_image_list[i].size()) # [240, 240]
+
+            break
+        # break
+
+
+   # print(augmented_image_list)
+
+
+    print('\n[+] Write on tensorboard')   
+    if writer:
+        for i, data in enumerate(augmented_image_list):
+            tag = 'img/' + str(i) 
+            writer.add_image(tag, data.view(1, 240, -1), global_step=0)
+            break
+
+    writer.close()      
+
+
+    # if writer:
+    #     for j in range():
+    #         tag = 'img/' + str(img_count) + '_' + str(j) 
+    #         # writer.add_image(tag,
+    #         #         concat_image_features(images[j], first[j]), global_step=step) 
+    #         # if j > 0:
+    #         #     fore = concat_image_features(fore, images[j])
+
+    #     writer.add_image(tag, fore, global_step=0)
+    #     img_count = img_count + 1
+
+    # writer.close()
+
+if __name__ == '__main__':
+    fire.Fire(eval)

+import pickle
+
+with open('logs/April_16_21:50:17__resnet50__None/augmentation.cp', 'rb') as f:
+    data = pickle.load(f)
+
+
+print(data)
+print(type(data))
\ No newline at end of file

@@ -4,7 +4,7 @@ hyperopt
 pillow==6.2.1
 natsort
 fire
-torch
-torchvision==0.4.1
+torchvision==0.2.2
+torch==1.1.0
 pandas
-sklearn
+sklearn
\ No newline at end of file

@@ -40,12 +40,12 @@ class TranslateXY(BaseTransform):
         return t(img)
 
 
-# class Rotate(BaseTransform):
+class Rotate(BaseTransform):
 
-#     def transform(self, img):
-#         degrees = self.mag * 360
-#         t = transforms.RandomRotation(degrees, Image.BILINEAR)
-#         return t(img)
+    def transform(self, img):
+        degrees = self.mag * 360
+        t = transforms.RandomRotation(degrees, Image.BILINEAR)
+        return t(img)
 
 
 class AutoContrast(BaseTransform):
@@ -55,10 +55,10 @@ class AutoContrast(BaseTransform):
         return ImageOps.autocontrast(img, cutoff=cutoff)
 
 
-# class Invert(BaseTransform):
+class Invert(BaseTransform):
 
-#     def transform(self, img):
-#         return ImageOps.invert(img)
+    def transform(self, img):
+        return ImageOps.invert(img)
 
 
 class Equalize(BaseTransform):
@@ -88,11 +88,11 @@ class Contrast(BaseTransform):
         return ImageEnhance.Contrast(img).enhance(factor)
 
 
-# class Color(BaseTransform):
+class Color(BaseTransform):
 
-#     def transform(self, img):
-#         factor = self.mag * 10
-#         return ImageEnhance.Color(img).enhance(factor)
+    def transform(self, img):
+        factor = self.mag * 10
+        return ImageEnhance.Color(img).enhance(factor)
 
 
 class Brightness(BaseTransform):
@@ -159,7 +159,10 @@ class CutoutOp(object):
         # print("\nnp.asarray(img) max: \n", np.amax(np.asarray(img)), np.asarray(img).shape) #(32, 32, 32)
         # img = Image.fromarray(mask*np.asarray(img)) #(32, 32, 32)
 
-        mask = np.reshape(mask, (32,32))
+        #mask = np.reshape(mask, (32, 32)) # (32, 32) -> (240, 240)
+
+        # getAugmented.py
+        mask = np.reshape(mask, (240, 240))
 
         #print("\n(img) max: \n", np.amax(np.asarray(img)), np.asarray(img).shape) #[0, 255] (32, 32)
         # print("\nmask: ", mask.shape)  #(32, 32)

@@ -59,11 +59,15 @@ def split_dataset(args, dataset, k):
     return Dm_indexes, Da_indexes
  
 
-#(images[j], first[j]), global_step=step)
+# concat_image_features(images[j], first[j])
 def concat_image_features(image, features, max_features=3):
     _, h, w = image.shape
+    #print("\nfsize: ", features.size()) # (1, 240, 240)
+   # features.size(0) = 64
+    #print(features.size(0))
+    #max_features = min(features.size(0), max_features)
 
-    max_features = min(features.size(0), max_features)
+    max_features = features.size(0)
     image_feature = image.clone()
 
     for i in range(max_features):
@@ -139,12 +143,12 @@ def parse_args(kwargs):
     kwargs['use_cuda'] =  kwargs['use_cuda'] if 'use_cuda' in kwargs else True
     kwargs['use_cuda'] =  kwargs['use_cuda'] and torch.cuda.is_available()
     kwargs['num_workers'] = kwargs['num_workers'] if 'num_workers' in kwargs else 4
-    kwargs['print_step'] = kwargs['print_step'] if 'print_step' in kwargs else 500
-    kwargs['val_step'] = kwargs['val_step'] if 'val_step' in kwargs else 500
+    kwargs['print_step'] = kwargs['print_step'] if 'print_step' in kwargs else 50
+    kwargs['val_step'] = kwargs['val_step'] if 'val_step' in kwargs else 50
     kwargs['scheduler'] = kwargs['scheduler'] if 'scheduler' in kwargs else 'exp'
-    kwargs['batch_size'] = kwargs['batch_size'] if 'batch_size' in kwargs else 64
+    kwargs['batch_size'] = kwargs['batch_size'] if 'batch_size' in kwargs else 8
     kwargs['start_step'] = kwargs['start_step'] if 'start_step' in kwargs else 0
-    kwargs['max_step'] = kwargs['max_step'] if 'max_step' in kwargs else 5000
+    kwargs['max_step'] = kwargs['max_step'] if 'max_step' in kwargs else 100
     kwargs['fast_auto_augment'] = kwargs['fast_auto_augment'] if 'fast_auto_augment' in kwargs else False
     kwargs['augment_path'] = kwargs['augment_path'] if 'augment_path' in kwargs else None
 
@@ -225,6 +229,7 @@ class CustomDataset(Dataset):
 
        if self.transform is not None:
             tensor_image = self.transform(image) ##
+
        return tensor_image, targets
 
 def get_dataset(args, transform, split='train'):
@@ -276,6 +281,14 @@ def get_dataloader(args, dataset, shuffle=False, pin_memory=True):
                                               pin_memory=pin_memory)
     return data_loader
 
+def get_aug_dataloader(args, dataset, shuffle=False, pin_memory=True):
+    data_loader = torch.utils.data.DataLoader(dataset,
+                                              # batch_size=args.batch_size,
+                                              shuffle=shuffle,
+                                              num_workers=args.num_workers,
+                                              pin_memory=pin_memory)
+    return data_loader
+
 
 def get_inf_dataloader(args, dataset):
     global current_epoch
@@ -304,7 +317,7 @@ def get_train_transform(args, model, log_dir=None):
             os.system('cp {} {}'.format(
                 args.augment_path, os.path.join(log_dir, 'augmentation.cp')))
         else:
-            transform = fast_auto_augment(args, model, K=4, B=1, num_process=4) ##
+            transform = fast_auto_augment(args, model, K=4, B=100, num_process=4) ##
             if log_dir:
                 cp.dump(transform, open(os.path.join(log_dir, 'augmentation.cp'), 'wb'))
 
@@ -436,7 +449,7 @@ def validate(args, model, criterion, valid_loader, step, writer, device=None):
             samples += images.size(0)
 
             if writer:
-               # print("\n3 images.size(0): ", images.size(0))
+                # print("\n images.size(0): ", images.size(0)) # batch size (last = n(imgs)%batch_size)
                 n_imgs = min(images.size(0), 10)
                 for j in range(n_imgs):
                     tag = 'valid/' + str(img_count)