# 初始化参数
train_img_path = r'E:\Datas\StreetCharsRecognition\mchar_train\*.png'
train_label_path = r'E:\Datas\StreetCharsRecognition\mchar_train.json'
val_img_path = r'E:\Datas\StreetCharsRecognition\mchar_val\*.png'
val_label_path = r'E:\Datas\StreetCharsRecognition\mchar_val.json'
test_img_path = r'E:\Datas\StreetCharsRecognition\mchar_test_a\*.png'
demo_submit_path = r'E:\Datas\StreetCharsRecognition\mchar_sample_submit_A.csv'
batch_size = 100
epochs = 20
lr = .001
is_predicting = False #默认is_predicting=False, 表明为训练过程
# 训练过程
if not is_predicting:
# 加载数据
train_path, train_label = Tools.dataFromPath(train_img_path,
train_label_path)
train_dataset = SVHNDataset(
train_path, train_label,
transforms.Compose([
transforms.Resize((64, 128)),
transforms.ColorJitter(0.3, 0.3, 0.2),
transforms.RandomRotation(5),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406],
[0.229, 0.224, 0.225])
]))
val_path, val_label = Tools.dataFromPath(val_img_path, val_label_path)
val_dataset = SVHNDataset(
val_path, val_label,
transforms.Compose([
transforms.Resize((64, 128)),
lbl = list(lbl) + (6 - len(lbl)) * [10]
return img, torch.from_numpy(np.array(lbl))
def __len__(self):
return len(self.img_path)
# Test SVHNDataset
if __name__ == '__main__':
# train_path = glob.glob(r'E:\Datas\StreetCharsRecognition\mchar_train\*.png')
# train_path.sort()
# train_json = json.load(open(r'E:\Datas\StreetCharsRecognition\mchar_train.json'))
# train_label = [train_json[x]['label'] for x in train_json]
train_path, train_label = Tools.dataFromPath(
r'E:\Datas\StreetCharsRecognition\mchar_train\*.png',
r'E:\Datas\StreetCharsRecognition\mchar_train.json')
print("扩增前数据集大小", ":", len(train_path))
train_loader = torch.utils.data.DataLoader(
SVHNDataset(
train_path, train_label,
transforms.Compose([
transforms.Resize((64, 128)),
transforms.ColorJitter(0.3, 0.3, 0.2),
transforms.RandomRotation(5),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406],
[0.229, 0.224, 0.225])
])),
