from COFGA_dataset import CofgaDataset
# ## Define run
# define name of output files
results_name = "ResNet50_no_aug__map.csv"
results_name_AP_val = "ResNet50_no_aug__AP_val.csv"
results_name_AP_train = "ResNet50_no_aug_AP_train.csv"
# ## Data augmentation
data_transform = None
# ## Loading the data
# loading the custom dataset
dataset = CofgaDataset(
csv_file='/zhome/b0/8/88043/COFGA_Project/dataset/train_preprocessed.csv',
root_dir='/zhome/b0/8/88043/COFGA_Project/dataset/root/train/resized/',
transform=data_transform)
print("Total number of images: ", len(dataset))
COFGA_headers = pd.read_csv(
'/zhome/b0/8/88043/COFGA_Project/dataset/train_preprocessed.csv')
COFGA_labels = COFGA_headers.columns.tolist()
COFGA_labels.pop(0)
COFGA_labels.insert(0, "epoch")
# ## Constructing trainLoader and validation loader
batch_size = 32
#performing vertical flip with a gven probability
prob = 0.8
vert_transform = transforms.Compose([
transforms.RandomVerticalFlip(prob),
transforms.ToTensor(),
])
# allows to chose randomly from the different transformations
transform_list = transforms.RandomChoice(
[rotation_transform, hoz_transform, vert_transform])
# ## Loading the data
# loading the custom dataset
dataset = CofgaDataset(csv_file='dataset/train_preprocessed.csv',
root_dir='dataset/root/train/resized/',
transform=transform_list)
print("Total number of images: ", len(dataset))
COFGA_headers = pd.read_csv('dataset/train_preprocessed.csv')
COFGA_labels = COFGA_headers.columns.tolist()
COFGA_labels.pop(0)
COFGA_labels.insert(0, "epoch")
# ## Constructing trainLoader and validation loader
batch_size = 32
# fraction of dataset to be validation set