mmocr/configs/textrecog/satrn/satrn_academic.py

_base_ = [
    '../../_base_/default_runtime.py',
    '../../_base_/schedules/schedule_adam_step_5e.py',
    '../../_base_/recog_models/satrn.py'
]

default_hooks = dict(logger=dict(type='LoggerHook', interval=50))

# dataset settings
dataset_type = 'OCRDataset'
data_root = 'tests/data/ocr_toy_dataset'
file_client_args = dict(backend='petrel')

model = dict(
    type='SATRN',
    backbone=dict(type='ShallowCNN', input_channels=3, hidden_dim=512),
    encoder=dict(
        type='SATRNEncoder',
        n_layers=12,
        n_head=8,
        d_k=512 // 8,
        d_v=512 // 8,
        d_model=512,
        n_position=100,
        d_inner=512 * 4,
        dropout=0.1),
    decoder=dict(
        type='NRTRDecoder',
        n_layers=6,
        d_embedding=512,
        n_head=8,
        d_model=512,
        d_inner=512 * 4,
        d_k=512 // 8,
        d_v=512 // 8,
        loss=dict(type='CELoss', flatten=True, ignore_first_char=True),
        max_seq_len=25,
        postprocessor=dict(type='AttentionPostprocessor')))

# optimizer
optim_wrapper = dict(type='OptimWrapper', optimizer=dict(type='Adam', lr=3e-4))

train_pipeline = [
    dict(type='LoadImageFromFile', file_client_args=file_client_args),
    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'))
]

# TODO Add Test Time Augmentation `MultiRotateAugOCR`
test_pipeline = [
    dict(type='LoadImageFromFile', file_client_args=file_client_args),
    dict(type='Resize', scale=(100, 32), keep_ratio=False),
    dict(
        type='PackTextRecogInputs',
        meta_keys=('img_path', 'ori_shape', 'img_shape', 'valid_ratio',
                   'instances'))
]

train_dataloader = dict(
    batch_size=64,
    num_workers=8,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_prefix=dict(img_path=None),
        ann_file='train_label.json',
        pipeline=train_pipeline))

val_dataloader = dict(
    batch_size=64,
    num_workers=4,
    persistent_workers=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_prefix=dict(img_path=None),
        ann_file='test_label.json',
        test_mode=True,
        pipeline=test_pipeline))
test_dataloader = val_dataloader

val_evaluator = [
    dict(
        type='WordMetric', mode=['exact', 'ignore_case',
                                 'ignore_case_symbol']),
    dict(type='CharMetric')
]
test_evaluator = val_evaluator
visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')