resolved classify acc err

조현아
Commit a5ce1660d3d53bce368d4a554286dc1f3586f168 a5ce1660 1 parent c4e3768c
Showing 3 changed files with 105 additions and 52 deletions
code/classifier/eval.py
code/classifier/train.py
code/classifier/utils.py
--- a/code/classifier/eval.py
View file @a5ce166
+++ b/code/classifier/eval.py
View file @a5ce166
@@ -5,19 +5,19 @@ from pprint import pprint
 import torch
 import torch.nn as nn
-from torch.utils.tensorboard import SummaryWriter
+import torchvision.transforms as transforms
+#from torch.utils.tensorboard import SummaryWriter
 from utils import *
 # command
 # python eval.py --model_path='logs/April_16_00:26:10__resnet50__None/'
-def eval(model_path, num_data):
+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)
-    args.batch_size = num_data
     pprint(args)
     device = torch.device('cuda' if args.use_cuda else 'cpu')
@@ -35,23 +35,25 @@ def eval(model_path, num_data):
     model.load_state_dict(torch.load(weight_path))
     print('\n[+] Load dataset')
-    test_dataset = get_dataset(args, 'test') 
+    transform = transforms.Compose([
+            transforms.Resize([240, 240]),
+            transforms.ToTensor() 
+    ])
+    test_dataset = get_dataset(args, transform, 'test') 
     test_loader = iter(get_dataloader(args, test_dataset)) ###
-    print('\n[+] Start testing')
+    # print('\n[+] Start testing')
-    writer = SummaryWriter(log_dir=model_path)
+    # writer = SummaryWriter(log_dir=model_path)
-    _test_res = validate(args, model, criterion, test_loader, step=0, writer=writer)
+    _test_res = validate(args, model, criterion, test_loader, step=0)
     print('\n[+] Valid results')
     print('  Acc@1 : {:.3f}%'.format(_test_res[0].data.cpu().numpy()[0]*100))
-    print('  Acc@5 : {:.3f}%'.format(_test_res[1].data.cpu().numpy()[0]*100))
+    print('  Loss : {:.3f}'.format(_test_res[1].data))
-    print('  Acc_all : {:.3f}%'.format(_test_res[2].data.cpu().numpy()[0]*100))
+    print('  Infer Time(per image) : {:.3f}ms'.format(_test_res[2]*1000 / len(test_dataset)))
-    print('  Loss : {:.3f}'.format(_test_res[3].data))
-    print('  Infer Time(per image) : {:.3f}ms'.format(_test_res[4]*1000 / len(test_dataset)))
-    writer.close()
+    #writer.close()
 if __name__ == '__main__':
     fire.Fire(eval)
--- a/code/classifier/train.py
View file @a5ce166
+++ b/code/classifier/train.py
View file @a5ce166
@@ -7,7 +7,7 @@ from pprint import pprint
 import torch.nn as nn
 import torch.backends.cudnn as cudnn
-from torch.utils.tensorboard import SummaryWriter
+#from torch.utils.tensorboard import SummaryWriter
 from networks import *
 from utils import *
@@ -27,7 +27,7 @@ def train(**kwargs):
     log_dir = os.path.join('/content/drive/My Drive/CD2 Project/runs/classify/', model_name)
     os.makedirs(os.path.join(log_dir, 'model'))
     json.dump(kwargs, open(os.path.join(log_dir, 'kwargs.json'), 'w'))
-    writer = SummaryWriter(log_dir=log_dir)
+    #writer = SummaryWriter(log_dir=log_dir)
     if args.seed is not None:
         random.seed(args.seed)
@@ -45,8 +45,10 @@ def train(**kwargs):
     #writer.add_graph(model)
     print('\n[+] Load dataset')
-    train_dataset = get_dataset(args, 'train')
+    transform = get_train_transform(args, model, log_dir) 
-    valid_dataset = get_dataset(args, 'val')
+    val_transform = get_valid_transform(args, model) 
+    train_dataset = get_dataset(args, transform, 'train')
+    valid_dataset = get_dataset(args, val_transform, 'val') 
     train_loader = iter(get_inf_dataloader(args, train_dataset))
     max_epoch = len(train_dataset) // args.batch_size
     best_acc = -1
@@ -62,16 +64,16 @@ def train(**kwargs):
     start_t = time.time()
     for step in range(args.start_step, args.max_step):
         batch = next(train_loader)
-        _train_res = train_step(args, model, optimizer, scheduler, criterion, batch, step, writer)
+        _train_res = train_step(args, model, optimizer, scheduler, criterion, batch, step)
         if step % args.print_step == 0:
             print('\n[+] Training step: {}/{}\tTraining epoch: {}/{}\tElapsed time: {:.2f}min\tLearning rate: {}'.format(
                 step, args.max_step, current_epoch, max_epoch, (time.time()-start_t)/60, optimizer.param_groups[0]['lr']))
-            writer.add_scalar('train/learning_rate', optimizer.param_groups[0]['lr'], global_step=step)
+            # writer.add_scalar('train/learning_rate', optimizer.param_groups[0]['lr'], global_step=step)
-            writer.add_scalar('train/acc1', _train_res[0], global_step=step)
+            # writer.add_scalar('train/acc1', _train_res[0], global_step=step)
-            writer.add_scalar('train/loss', _train_res[1], global_step=step)
+            # writer.add_scalar('train/loss', _train_res[1], global_step=step)
-            writer.add_scalar('train/forward_time', _train_res[2], global_step=step)
+            # writer.add_scalar('train/forward_time', _train_res[2], global_step=step)
-            writer.add_scalar('train/backward_time', _train_res[3], global_step=step)
+            # writer.add_scalar('train/backward_time', _train_res[3], global_step=step)
             print('  Acc@1 : {:.3f}%'.format(_train_res[0].data.cpu().numpy()[0]*100))
             print('  Loss : {}'.format(_train_res[1].data))
             print('  FW Time : {:.3f}ms'.format(_train_res[2]*1000))
@@ -80,10 +82,10 @@ def train(**kwargs):
         if step % args.val_step == args.val_step-1:
             # print("\nstep, args.val_step: ", step, args.val_step)
             valid_loader = iter(get_dataloader(args, valid_dataset))
-            _valid_res = validate(args, model, criterion, valid_loader, step, writer)
+            _valid_res = validate(args, model, criterion, valid_loader, step)
-            print('\n[+] Valid results')
+            print('\n[+] (Valid results) Valid step: {}/{}'.format(step, args.max_step))
-            writer.add_scalar('valid/acc1', _valid_res[0], global_step=step)
+            # writer.add_scalar('valid/acc1', _valid_res[0], global_step=step)
-            writer.add_scalar('valid/loss', _valid_res[1], global_step=step)
+            # writer.add_scalar('valid/loss', _valid_res[1], global_step=step)
             print('  Acc@1 : {:.3f}%'.format(_valid_res[0].data.cpu().numpy()[0]*100))
             print('  Loss : {}'.format(_valid_res[1].data))
@@ -92,7 +94,7 @@ def train(**kwargs):
                 torch.save(model.state_dict(), os.path.join(log_dir, "model","model.pt"))
                 print('\n[+] Model saved')
-    writer.close()
+   # writer.close()
 if __name__ == '__main__':
--- a/code/classifier/utils.py
View file @a5ce166
+++ b/code/classifier/utils.py
View file @a5ce166
@@ -23,6 +23,7 @@ from sklearn.model_selection import KFold
 from networks import basenet, grayResNet2
+DATASET_PATH = '/content/drive/My Drive/CD2 Project/'
 TRAIN_DATASET_PATH = '/content/drive/My Drive/CD2 Project/data/nonaug+Normal_train/'
 TRAIN_TARGET_PATH = '/content/drive/My Drive/CD2 Project/data/train_nonaug_classify_target.csv'
@@ -131,17 +132,17 @@ def parse_args(kwargs):
     kwargs['dataset'] =  kwargs['dataset'] if 'dataset' in kwargs else 'BraTS'
     kwargs['network'] =  kwargs['network'] if 'network' in kwargs else 'resnet50'
     kwargs['optimizer'] = kwargs['optimizer'] if 'optimizer' in kwargs else 'adam'
-    kwargs['learning_rate'] = kwargs['learning_rate'] if 'learning_rate' in kwargs else 0.001
+    kwargs['learning_rate'] = kwargs['learning_rate'] if 'learning_rate' in kwargs else 0.01
     kwargs['seed'] =  kwargs['seed'] if 'seed' in kwargs else None
     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 100
+    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 100
+    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 32
+    kwargs['batch_size'] = kwargs['batch_size'] if 'batch_size' in kwargs else 16
     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 2500
+    kwargs['max_step'] = kwargs['max_step'] if 'max_step' in kwargs else 500
     kwargs['augment_path'] = kwargs['augment_path'] if 'augment_path' in kwargs else None
     # to named tuple
@@ -155,11 +156,10 @@ def select_model(args):
     'resnet50':grayResNet2.resnet50(), 'resnet101':grayResNet2.resnet101(), 'resnet152':grayResNet2.resnet152()}
     if args.network in resnet_dict:
-        backbone = resnet_dict[args.network]
+        model = resnet_dict[args.network]
-        model = basenet.BaseNet(backbone, args)
+    # else: # 3 channels
-    else:
+    # backbone = models.__dict__[args.network]()
-        Net = getattr(importlib.import_module('networks.{}'.format(args.network)), 'Net')
+    # model = basenet.BaseNet(backbone, args)
-        model = Net(args)
     #print(model) # print model architecture
     return model
@@ -187,16 +187,44 @@ def select_scheduler(args, optimizer):
     else:
         raise Exception('Unknown Scheduler')
+def get_train_transform(args, model, transform, log_dir=None):
+    if args.dataset == 'cifar10':
+        transform = transforms.Compose([
+            transforms.Pad(4),
+            transforms.RandomCrop(32),
+            transforms.RandomHorizontalFlip(),
+            transforms.ToTensor()
+        ])
+
+    else:
+        transform = transforms.Compose([
+            transforms.Resize([240, 240]),
+            transforms.ToTensor() 
+    ])
+
+    return transform
+
+def get_valid_transform(args, model):
+    if args.dataset == 'cifar10':
+        val_transform = transforms.Compose([
+            transforms.Resize(32),
+            transforms.ToTensor()
+        ])
+
+    else:
+        val_transform = transforms.Compose([
+            transforms.Resize([240, 240]),
+            transforms.ToTensor() 
+    ])
+
+    return val_transform
 class CustomDataset(Dataset): 
-  def __init__(self, data_path, csv_path):
+  def __init__(self, data_path, csv_path, transform):
     self.path = data_path
     self.imgs = natsorted(os.listdir(data_path))
     self.len = len(self.imgs)
-    self.transform = transforms.Compose([
+    self.transform = transform
-        transforms.Resize([240, 240]),
-        transforms.ToTensor() 
-    ])
     df = pd.read_csv(csv_path)
     targets_list = []
@@ -215,6 +243,7 @@ class CustomDataset(Dataset):
     targets = self.targets[idx]
     image = Image.open(img_loc)
     image = self.transform(image)
+    #print("\n idx, img, targets: ", idx, img_loc, targets)
     return image, targets
@@ -222,12 +251,32 @@ class CustomDataset(Dataset):
 def get_dataset(args, transform, split='train'):
     assert split in ['train', 'val', 'test']
-    if split in ['train']:
+    if args.dataset == 'cifar10':
-        dataset = CustomDataset(TRAIN_DATASET_PATH, TRAIN_TARGET_PATH)
+        train = split in ['train', 'val', 'trainval']
-    elif split in ['val']: 
+        dataset = torchvision.datasets.CIFAR10(DATASET_PATH,
-        dataset = CustomDataset(VAL_DATASET_PATH, VAL_TARGET_PATH)
+                                               train=train,
-    else : #  test
+                                               transform=transform,
-        dataset = CustomDataset(TEST_DATASET_PATH, TEST_TARGET_PATH)
+                                               download=True)
+
+        if split in ['train', 'val']:
+            split_path = os.path.join(DATASET_PATH,
+                    'cifar-10-batches-py', 'train_val_index.cp')
+
+            if not os.path.exists(split_path):
+                [train_index], [val_index] = split_dataset(args, dataset, k=1)
+                split_index = {'train':train_index, 'val':val_index}
+                cp.dump(split_index, open(split_path, 'wb'))
+
+            split_index = cp.load(open(split_path, 'rb'))
+            dataset = Subset(dataset, split_index[split])
+
+    else:                                            
+        if split in ['train']:
+            dataset = CustomDataset(TRAIN_DATASET_PATH, TRAIN_TARGET_PATH, transform)
+        elif split in ['val']: 
+            dataset = CustomDataset(VAL_DATASET_PATH, VAL_TARGET_PATH, transform)
+        else : #  test
+            dataset = CustomDataset(TEST_DATASET_PATH, TEST_TARGET_PATH, transform)
     return dataset
@@ -261,7 +310,7 @@ def get_inf_dataloader(args, dataset):
-def train_step(args, model, optimizer, scheduler, criterion, batch, step, writer, device=None):
+def train_step(args, model, optimizer, scheduler, criterion, batch, step, device=None):
     model.train()
     images, target = batch
@@ -275,7 +324,7 @@ def train_step(args, model, optimizer, scheduler, criterion, batch, step, writer
     # compute output
     start_t = time.time()
-    output, first = model(images)
+    output= model(images)
     forward_t = time.time() - start_t
     loss = criterion(output, target)
@@ -323,7 +372,7 @@ def accuracy(output, target, topk=(1,)):
             res.append(correct_k)
         return res
-def validate(args, model, criterion, valid_loader, step, writer, device=None):
+def validate(args, model, criterion, valid_loader, step, device=None):
     # switch to evaluate mode
     model.eval()
@@ -344,7 +393,7 @@ def validate(args, model, criterion, valid_loader, step, writer, device=None):
                 target = target.cuda(non_blocking=True)
             # compute output
-            output, first = model(images)
+            output  = model(images)
             loss = criterion(output, target)
             infer_t += time.time() - start_t