Repeating characters in a word are being suppressed by SATRN_small model

[raushani2v]

Hi,

I have trained a SATRN_small model on custom dataset for character recognition. The repeating characters in a word are being suppressed by the model. for e.g, 40799932 is predicted as 4079932.

What is the reason for this problem? Is there any solution?

I am using following configuration:

default_scope = 'mmocr'
env_cfg = dict(
    cudnn_benchmark=False,
    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
    dist_cfg=dict(backend='nccl'))
randomness = dict(seed=None)
default_hooks = dict(
    timer=dict(type='IterTimerHook'),
    logger=dict(type='LoggerHook', interval=10),
    param_scheduler=dict(type='ParamSchedulerHook'),
    checkpoint=dict(type='CheckpointHook', interval=50, save_best='auto'),
    sampler_seed=dict(type='DistSamplerSeedHook'),
    sync_buffer=dict(type='SyncBuffersHook'),
    visualization=dict(
        type='VisualizationHook',
        interval=1,
        enable=False,
        show=False,
        draw_gt=False,
        draw_pred=False))
log_level = 'INFO'
log_processor = dict(type='LogProcessor', window_size=10, by_epoch=True)
load_from = None
resume = False
val_evaluator = dict(
    type='MultiDatasetsEvaluator',
    metrics=[
        dict(
            type='WordMetric',
            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
        dict(type='CharMetric')
    ],
    dataset_prefixes=['custom_dataset'])
test_evaluator = dict(
    type='MultiDatasetsEvaluator',
    metrics=[
        dict(
            type='WordMetric',
            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
        dict(type='CharMetric')
    ],
    dataset_prefixes=['custom_dataset'])
vis_backends = [dict(type='LocalVisBackend')]
visualizer = dict(
    type='TextRecogLocalVisualizer',
    name='visualizer',
    vis_backends=[dict(type='LocalVisBackend')])
tta_model = dict(type='EncoderDecoderRecognizerTTAModel')
train_dataset_type = 'OCRDataset'
train_img_prefix = './data/wagon_training_data_05-09-2023/images'
train_anno_file = './data/wagon_training_data_05-09-2023/annotation_train.json'
train = dict(
    type='OCRDataset',
    data_root=
    './data/wagon_training_data_05-09-2023/images',
    ann_file=
    './data/wagon_training_data_05-09-2023/annotation_train.json',
    pipeline=None,
    test_mode=False)
train_list = [
    dict(
        type='OCRDataset',
        data_root=
        './data/wagon_training_data_05-09-2023/images',
        ann_file=
        './data/wagon_training_data_05-09-2023/annotation_train.json',
        pipeline=None,
        test_mode=False)
]
test_dataset_type = 'OCRDataset'
test_img_prefix = './data/wagon_training_data_05-09-2023/images'
test_anno_file = './data/wagon_training_data_05-09-2023/annotation_val.json'
test = dict(
    type='OCRDataset',
    data_root=
    './data/wagon_training_data_05-09-2023/images',
    ann_file=
    './data/wagon_training_data_05-09-2023/annotation_val.json',
    pipeline=None,
    test_mode=True)
test_list = [
    dict(
        type='OCRDataset',
        data_root=
        './data/wagon_training_data_05-09-2023/images',
        ann_file=
        './data/wagon_training_data_05-09-2023/annotation_val.json',
        pipeline=None,
        test_mode=True)
]
optim_wrapper = dict(
    type='OptimWrapper', optimizer=dict(type='Adam', lr=0.0003))
train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=100, val_interval=1)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
param_scheduler = [dict(type='MultiStepLR', milestones=[60, 80], end=100)]
dictionary = dict(
    type='Dictionary',
    dict_file=
    './configs/textrecog/satrn/../../../dicts/digits.txt',
    with_padding=True,
    with_unknown=True,
    same_start_end=True,
    with_start=True,
    with_end=True)
model = dict(
    type='SATRN',
    backbone=dict(type='ShallowCNN', input_channels=3, hidden_dim=256),
    encoder=dict(
        type='SATRNEncoder',
        n_layers=6,
        n_head=8,
        d_k=32,
        d_v=32,
        d_model=256,
        n_position=100,
        d_inner=1024,
        dropout=0.1),
    decoder=dict(
        type='NRTRDecoder',
        n_layers=6,
        d_embedding=256,
        n_head=8,
        d_model=256,
        d_inner=1024,
        d_k=32,
        d_v=32,
        module_loss=dict(
            type='CEModuleLoss', flatten=True, ignore_first_char=True),
        dictionary=dict(
            type='Dictionary',
            dict_file=
            './configs/textrecog/satrn/../../../dicts/digits.txt',
            with_padding=True,
            with_unknown=True,
            same_start_end=True,
            with_start=True,
            with_end=True),
        max_seq_len=25,
        postprocessor=dict(type='AttentionPostprocessor')),
    data_preprocessor=dict(
        type='TextRecogDataPreprocessor',
        mean=[123.675, 116.28, 103.53],
        std=[58.395, 57.12, 57.375]))
train_pipeline = [
    dict(type='LoadImageFromFile', ignore_empty=True, min_size=2),
    dict(type='LoadOCRAnnotations', with_text=True),
    dict(type='Resize', scale=(100, 32), keep_ratio=False),
    dict(
        type='PackTextRecogInputs',
        meta_keys=('img_path', 'ori_shape', 'img_shape', 'valid_ratio'))
]
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='Resize', scale=(100, 32), keep_ratio=False),
    dict(type='LoadOCRAnnotations', with_text=True),
    dict(
        type='PackTextRecogInputs',
        meta_keys=('img_path', 'ori_shape', 'img_shape', 'valid_ratio'))
]
tta_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(
        type='TestTimeAug',
        transforms=[[{
            'type':
            'ConditionApply',
            'true_transforms': [{
                'type':
                'ImgAugWrapper',
                'args': [{
                    'cls': 'Rot90',
                    'k': 0,
                    'keep_size': False
                }]
            }],
            'condition':
            "results['img_shape'][1]<results['img_shape'][0]"
        }, {
            'type':
            'ConditionApply',
            'true_transforms': [{
                'type':
                'ImgAugWrapper',
                'args': [{
                    'cls': 'Rot90',
                    'k': 1,
                    'keep_size': False
                }]
            }],
            'condition':
            "results['img_shape'][1]<results['img_shape'][0]"
        }, {
            'type':
            'ConditionApply',
            'true_transforms': [{
                'type':
                'ImgAugWrapper',
                'args': [{
                    'cls': 'Rot90',
                    'k': 3,
                    'keep_size': False
                }]
            }],
            'condition':
            "results['img_shape'][1]<results['img_shape'][0]"
        }], [{
            'type': 'Resize',
            'scale': (100, 32),
            'keep_ratio': False
        }], [{
            'type': 'LoadOCRAnnotations',
            'with_text': True
        }],
                    [{
                        'type':
                        'PackTextRecogInputs',
                        'meta_keys':
                        ('img_path', 'ori_shape', 'img_shape', 'valid_ratio')
                    }]])
]
train_dataset = dict(
    type='ConcatDataset',
    datasets=[
        dict(
            type='OCRDataset',
            data_root=
            './data/wagon_training_data_05-09-2023/images',
            ann_file=
            './data/wagon_training_data_05-09-2023/annotation_train.json',
            pipeline=None,
            test_mode=False)
    ],
    pipeline=[
        dict(type='LoadImageFromFile', ignore_empty=True, min_size=2),
        dict(type='LoadOCRAnnotations', with_text=True),
        dict(type='Resize', scale=(100, 32), keep_ratio=False),
        dict(
            type='PackTextRecogInputs',
            meta_keys=('img_path', 'ori_shape', 'img_shape', 'valid_ratio'))
    ])
test_dataset = dict(
    type='ConcatDataset',
    datasets=[
        dict(
            type='OCRDataset',
            data_root=
            './data/wagon_training_data_05-09-2023/images',
            ann_file=
            './data/wagon_training_data_05-09-2023/annotation_val.json',
            pipeline=None,
            test_mode=True)
    ],
    pipeline=[
        dict(type='LoadImageFromFile'),
        dict(type='Resize', scale=(100, 32), keep_ratio=False),
        dict(type='LoadOCRAnnotations', with_text=True),
        dict(
            type='PackTextRecogInputs',
            meta_keys=('img_path', 'ori_shape', 'img_shape', 'valid_ratio'))
    ])
train_dataloader = dict(
    batch_size=256,
    num_workers=4,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type='ConcatDataset',
        datasets=[
            dict(
                type='OCRDataset',
                data_root=
                './data/wagon_training_data_05-09-2023/images',
                ann_file=
                './data/wagon_training_data_05-09-2023/annotation_train.json',
                pipeline=None,
                test_mode=False)
        ],
        pipeline=[
            dict(type='LoadImageFromFile', ignore_empty=True, min_size=2),
            dict(type='LoadOCRAnnotations', with_text=True),
            dict(type='Resize', scale=(100, 32), keep_ratio=False),
            dict(
                type='PackTextRecogInputs',
                meta_keys=('img_path', 'ori_shape', 'img_shape',
                           'valid_ratio'))
        ]))
test_dataloader = dict(
    batch_size=256,
    num_workers=4,
    persistent_workers=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type='ConcatDataset',
        datasets=[
            dict(
                type='OCRDataset',
                data_root=
                './data/wagon_training_data_05-09-2023/images',
                ann_file=
                './data/wagon_training_data_05-09-2023/annotation_val.json',
                pipeline=None,
                test_mode=True)
        ],
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(type='Resize', scale=(100, 32), keep_ratio=False),
            dict(type='LoadOCRAnnotations', with_text=True),
            dict(
                type='PackTextRecogInputs',
                meta_keys=('img_path', 'ori_shape', 'img_shape',
                           'valid_ratio'))
        ]))
val_dataloader = dict(
    batch_size=256,
    num_workers=4,
    persistent_workers=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type='ConcatDataset',
        datasets=[
            dict(
                type='OCRDataset',
                data_root=
                './data/wagon_training_data_05-09-2023/images',
                ann_file=
                './data/wagon_training_data_05-09-2023/annotation_val.json',
                pipeline=None,
                test_mode=True)
        ],
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(type='Resize', scale=(100, 32), keep_ratio=False),
            dict(type='LoadOCRAnnotations', with_text=True),
            dict(
                type='PackTextRecogInputs',
                meta_keys=('img_path', 'ori_shape', 'img_shape',
                           'valid_ratio'))
        ]))
auto_scale_lr = dict(base_batch_size=512)
launcher = 'none'
work_dir = './work_dirs/satrn_small_custom_dataset'