[Config] recog dataset config refactor (#1275)

configs/_base_/recog_datasets/MJ_train.py

@@ -1,12 +0,0 @@
-# Text Recognition Training set, including:
-# Synthetic Datasets: Syn90k
-
-mj_rec_train = dict(
-    type='OCRDataset',
-    data_root='data/rec',
-    data_prefix=dict(img_path='Syn90k/mnt/ramdisk/max/90kDICT32px'),
-    ann_file='Syn90k/train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-train_list = [mj_rec_train]

configs/_base_/recog_datasets/ST_MJ_alphanumeric_train.py

@@ -1,24 +0,0 @@
-# Text Recognition Training set, including:
-# Synthetic Datasets: SynthText, Syn90k
-# Both annotations are filtered so that
-# only alphanumeric terms are left
-data_root = 'data/rec'
-
-mj_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path='Syn90k/mnt/ramdisk/max/90kDICT32px'),
-    ann_file='Syn90k/train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-st_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(
-        img_path='SynthText/synthtext/SynthText_patch_horizontal'),
-    ann_file='SynthText/alphanumeric_train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-train_list = [mj_rec_train, st_rec_train]

configs/_base_/recog_datasets/ST_MJ_train.py

@@ -1,23 +0,0 @@
-# Text Recognition Training set, including:
-# Synthetic Datasets: SynthText, Syn90k
-
-data_root = 'data/rec'
-
-mj_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path='Syn90k/mnt/ramdisk/max/90kDICT32px'),
-    ann_file='Syn90k/train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-st_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(
-        img_path='SynthText/synthtext/SynthText_patch_horizontal'),
-    ann_file='SynthText/train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-train_list = [mj_rec_train, st_rec_train]

configs/_base_/recog_datasets/ST_SA_MJ_real_train.py

@@ -1,91 +0,0 @@
-# Text Recognition Training set, including:
-# Synthetic Datasets: SynthText, SynthAdd, Syn90k
-# Real Dataset: IC11, IC13, IC15, COCO-Test, IIIT5k
-data_root = 'data/rec'
-
-train_img_prefix1 = 'icdar_2011'
-train_img_prefix2 = 'icdar_2013'
-train_img_prefix3 = 'icdar_2015'
-train_img_prefix4 = 'coco_text_v1'
-train_img_prefix5 = 'IIIT5K'
-train_img_prefix6 = 'synthtext_add'
-train_img_prefix7 = 'SynthText/synthtext/SynthText_patch_horizontal'
-train_img_prefix8 = 'Syn90k/mnt/ramdisk/max/90kDICT32px'
-
-train_ann_file1 = 'icdar_2011/train_labels.json'
-train_ann_file2 = 'icdar_2013/train_labels.json'
-train_ann_file3 = 'icdar_2015/train_labels.json'
-train_ann_file4 = 'coco_text_v1/train_labels.json'
-train_ann_file5 = 'IIIT5K/train_labels.json'
-train_ann_file6 = 'synthtext_add/train_labels.json'
-train_ann_file7 = 'SynthText/shuffle_train_labels.json'
-train_ann_file8 = 'Syn90k/shuffle_train_labels.json'
-
-ic11_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix1),
-    ann_file=train_ann_file1,
-    test_mode=False,
-    pipeline=None)
-
-ic13_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix2),
-    ann_file=train_ann_file2,
-    test_mode=False,
-    pipeline=None)
-
-ic15_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix3),
-    ann_file=train_ann_file3,
-    test_mode=False,
-    pipeline=None)
-
-cocov1_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix4),
-    ann_file=train_ann_file4,
-    test_mode=False,
-    pipeline=None)
-
-iiit5k_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix5),
-    ann_file=train_ann_file5,
-    test_mode=False,
-    pipeline=None)
-
-st_add_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix6),
-    ann_file=train_ann_file6,
-    test_mode=False,
-    pipeline=None)
-
-st_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix7),
-    ann_file=train_ann_file7,
-    test_mode=False,
-    pipeline=None)
-
-mj_rec_trian = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path=train_img_prefix8),
-    ann_file=train_ann_file8,
-    test_mode=False,
-    pipeline=None)
-
-train_list = [
-    ic13_rec_train, ic11_rec_train, ic15_rec_train, cocov1_rec_train,
-    iiit5k_rec_train, st_add_rec_train, st_rec_train, mj_rec_trian
-]

configs/_base_/recog_datasets/ST_SA_MJ_train.py

@@ -1,30 +0,0 @@
-# Text Recognition Training set, including:
-# Synthetic Datasets: SynthText, Syn90k
-data_root = 'data/rec'
-
-mj_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path='Syn90k/mnt/ramdisk/max/90kDICT32px'),
-    ann_file='Syn90k/train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-st_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(
-        img_path='SynthText/synthtext/SynthText_patch_horizontal'),
-    ann_file='SynthText/train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-st_add_rec_train = dict(
-    type='OCRDataset',
-    data_root=data_root,
-    data_prefix=dict(img_path='synthtext_add'),
-    ann_file='synthtext_add/train_labels.json',
-    test_mode=False,
-    pipeline=None)
-
-train_list = [mj_rec_train, st_rec_train, st_add_rec_train]

configs/_base_/recog_datasets/academic_test.py

@@ -1,170 +0,0 @@
-# Text Recognition Testing set, including:
-# Regular Datasets: IIIT5K, SVT, IC13
-# Irregular Datasets: IC15, SVTP, CT80
-
-test_root = 'data/rec'
-
-test_img_prefix1 = 'IIIT5K/'
-test_img_prefix2 = 'svt/'
-test_img_prefix3 = 'icdar_2013/Challenge2_Test_Task3_Images/'
-test_img_prefix4 = 'icdar_2015/ch4_test_word_images_gt'
-test_img_prefix5 = 'svtp/'
-test_img_prefix6 = 'ct80/'
-
-test_ann_file1 = 'IIIT5K/test_labels.json'
-test_ann_file2 = 'svt/test_labels.json'
-test_ann_file3 = 'icdar_2013/test_labels.json'
-test_ann_file4 = 'icdar_2015/test_labels.json'
-test_ann_file5 = 'svtp/test_labels.json'
-test_ann_file6 = 'ct80/test_labels.json'
-
-iiit5k_rec_test = dict(
-    type='OCRDataset',
-    data_root=test_root,
-    data_prefix=dict(img_path=test_img_prefix1),
-    ann_file=test_ann_file1,
-    test_mode=True,
-    pipeline=None)
-
-svt_rec_test = dict(
-    type='OCRDataset',
-    data_root=test_root,
-    data_prefix=dict(img_path=test_img_prefix2),
-    ann_file=test_ann_file2,
-    test_mode=True,
-    pipeline=None)
-
-ic13_rec_test = dict(
-    type='OCRDataset',
-    data_root=test_root,
-    data_prefix=dict(img_path=test_img_prefix3),
-    ann_file=test_ann_file3,
-    test_mode=True,
-    pipeline=None)
-
-ic15_rec_test = dict(
-    type='OCRDataset',
-    data_root=test_root,
-    data_prefix=dict(img_path=test_img_prefix4),
-    ann_file=test_ann_file4,
-    test_mode=True,
-    pipeline=None)
-
-svtp_rec_test = dict(
-    type='OCRDataset',
-    data_root=test_root,
-    data_prefix=dict(img_path=test_img_prefix5),
-    ann_file=test_ann_file5,
-    test_mode=True,
-    pipeline=None)
-
-cute80_rec_test = dict(
-    type='OCRDataset',
-    data_root=test_root,
-    data_prefix=dict(img_path=test_img_prefix6),
-    ann_file=test_ann_file6,
-    test_mode=True,
-    pipeline=None)
-
-iiit5k_val_dataloader = dict(
-    batch_size=1,
-    num_workers=4,
-    persistent_workers=True,
-    drop_last=False,
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    dataset=iiit5k_rec_test)
-
-svt_val_dataloader = dict(
-    batch_size=1,
-    num_workers=4,
-    persistent_workers=True,
-    drop_last=False,
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    dataset=svt_rec_test)
-
-ic13_val_dataloader = dict(
-    batch_size=1,
-    num_workers=4,
-    persistent_workers=True,
-    drop_last=False,
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    dataset=ic13_rec_test)
-
-ic15_val_dataloader = dict(
-    batch_size=1,
-    num_workers=4,
-    persistent_workers=True,
-    drop_last=False,
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    dataset=ic15_rec_test)
-
-svtp_val_dataloader = dict(
-    batch_size=1,
-    num_workers=4,
-    persistent_workers=True,
-    drop_last=False,
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    dataset=svtp_rec_test)
-
-cute80_val_dataloader = dict(
-    batch_size=1,
-    num_workers=4,
-    persistent_workers=True,
-    drop_last=False,
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    dataset=cute80_rec_test)
-
-val_dataloader = [
-    iiit5k_val_dataloader, svt_val_dataloader, ic13_val_dataloader,
-    ic15_val_dataloader, svtp_val_dataloader, cute80_val_dataloader
-]
-
-test_dataloader = val_dataloader
-
-val_evaluator = [[
-    dict(
-        type='WordMetric',
-        mode=['exact', 'ignore_case', 'ignore_case_symbol'],
-        prefix='IIIT5K'),
-    dict(type='CharMetric', prefix='IIIT5K')
-],
-                 [
-                     dict(
-                         type='WordMetric',
-                         mode=['exact', 'ignore_case', 'ignore_case_symbol'],
-                         prefix='svt'),
-                     dict(type='CharMetric', prefix='svt')
-                 ],
-                 [
-                     dict(
-                         type='WordMetric',
-                         mode=['exact', 'ignore_case', 'ignore_case_symbol'],
-                         prefix='icdar_2013'),
-                     dict(type='CharMetric', prefix='icdar_2013')
-                 ],
-                 [
-                     dict(
-                         type='WordMetric',
-                         mode=['exact', 'ignore_case', 'ignore_case_symbol'],
-                         prefix='icdar_2015'),
-                     dict(type='CharMetric', prefix='icdar_2015')
-                 ],
-                 [
-                     dict(
-                         type='WordMetric',
-                         mode=['exact', 'ignore_case', 'ignore_case_symbol'],
-                         prefix='svtp'),
-                     dict(type='CharMetric', prefix='svtp')
-                 ],
-                 [
-                     dict(
-                         type='WordMetric',
-                         mode=['exact', 'ignore_case', 'ignore_case_symbol'],
-                         prefix='ct80'),
-                     dict(type='CharMetric', prefix='ct80')
-                 ]]
-test_evaluator = val_evaluator
-test_list = [
-    iiit5k_rec_test, svt_rec_test, ic13_rec_test, ic15_rec_test, svtp_rec_test,
-    cute80_rec_test
-]

configs/_base_/recog_datasets/coco_text_v1.py

@@ -0,0 +1,8 @@
+cocov1_rec_train_data_root = 'data/rec/coco_text_v1'
+
+cocov1_rec_train = dict(
+    type='OCRDataset',
+    data_root=cocov1_rec_train_data_root,
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)

configs/_base_/recog_datasets/cute80.py

@@ -0,0 +1,8 @@
+cute80_rec_data_root = 'data/rec/ct80/'
+
+cute80_rec_test = dict(
+    type='OCRDataset',
+    data_root=cute80_rec_data_root,
+    ann_file='test_labels.json',
+    test_mode=True,
+    pipeline=None)

configs/_base_/recog_datasets/icdar2011.py

@@ -0,0 +1,8 @@
+ic11_rec_data_root = 'data/rec/icdar_2011/'
+
+ic11_rec_train = dict(
+    type='OCRDataset',
+    data_root=ic11_rec_data_root,
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)

configs/_base_/recog_datasets/icdar2013.py

@@ -0,0 +1,16 @@
+ic13_rec_data_root = 'data/rec/icdar_2013/'
+
+ic13_rec_train = dict(
+    type='OCRDataset',
+    data_root=ic13_rec_data_root,
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)
+
+ic13_rec_test = dict(
+    type='OCRDataset',
+    data_root=ic13_rec_data_root,
+    data_prefix=dict(img_path='Challenge2_Test_Task3_Images/'),
+    ann_file='test_labels.json',
+    test_mode=True,
+    pipeline=None)

configs/_base_/recog_datasets/icdar2015.py

@@ -0,0 +1,16 @@
+ic15_rec_data_root = 'data/rec/icdar_2015/'
+
+ic15_rec_train = dict(
+    type='OCRDataset',
+    data_root=ic15_rec_data_root,
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)
+
+ic15_rec_test = dict(
+    type='OCRDataset',
+    data_root=ic15_rec_data_root,
+    data_prefix=dict(img_path='ch4_test_word_images_gt/'),
+    ann_file='test_labels.json',
+    test_mode=True,
+    pipeline=None)

configs/_base_/recog_datasets/iiit5k.py

@@ -0,0 +1,15 @@
+iiit5k_rec_data_root = 'data/rec/IIIT5K/'
+
+iiit5k_rec_train = dict(
+    type='OCRDataset',
+    data_root=iiit5k_rec_data_root,
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)
+
+iiit5k_rec_test = dict(
+    type='OCRDataset',
+    data_root=iiit5k_rec_data_root,
+    ann_file='test_labels.json',
+    test_mode=True,
+    pipeline=None)

configs/_base_/recog_datasets/mjsynth.py

@@ -0,0 +1,17 @@
+mj_rec_data_root = 'data/rec/'
+
+mj_rec_train = dict(
+    type='OCRDataset',
+    data_root=mj_rec_data_root,
+    data_prefix=dict(img_path='mnt/ramdisk/max/90kDICT32px'),
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)
+
+mj_sub_rec_train = dict(
+    type='OCRDataset',
+    data_root=mj_rec_data_root,
+    data_prefix=dict(img_path='mnt/ramdisk/max/90kDICT32px'),
+    ann_file='subset_train_labels.json',
+    test_mode=False,
+    pipeline=None)

configs/_base_/recog_datasets/svt.py

@@ -0,0 +1,8 @@
+svt_rec_data_root = 'data/rec/svt/'
+
+svt_rec_test = dict(
+    type='OCRDataset',
+    data_root=svt_rec_data_root,
+    ann_file='test_labels.json',
+    test_mode=True,
+    pipeline=None)

configs/_base_/recog_datasets/svtp.py

@@ -0,0 +1,8 @@
+svtp_rec_data_root = 'data/rec/svtp/'
+
+svtp_rec_test = dict(
+    type='OCRDataset',
+    data_root=svtp_rec_data_root,
+    ann_file='test_labels.json',
+    test_mode=True,
+    pipeline=None)

configs/_base_/recog_datasets/synthtext.py

@@ -0,0 +1,25 @@
+st_data_root = 'data/rec/SynthText/'
+
+st_rec_train = dict(
+    type='OCRDataset',
+    data_root=st_data_root,
+    data_prefix=dict(img_path='synthtext/SynthText_patch_horizontal'),
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)
+
+st_an_rec_train = dict(
+    type='OCRDataset',
+    data_root=st_data_root,
+    data_prefix=dict(img_path='synthtext/SynthText_patch_horizontal'),
+    ann_file='alphanumeric_train_labels.json',
+    test_mode=False,
+    pipeline=None)
+
+st_sub_rec_train = dict(
+    type='OCRDataset',
+    data_root=st_data_root,
+    data_prefix=dict(img_path='synthtext/SynthText_patch_horizontal'),
+    ann_file='subset_train_labels.json',
+    test_mode=False,
+    pipeline=None)

configs/_base_/recog_datasets/synthtext_add.py

@@ -0,0 +1,8 @@
+st_add_rec_data_root = 'data/rec/st_add/'
+
+st_add_rec_train = dict(
+    type='OCRDataset',
+    data_root=st_add_rec_data_root,
+    ann_file='train_labels.json',
+    test_mode=False,
+    pipeline=None)

configs/textrecog/abinet/base.py

@@ -1,12 +1,22 @@
 _base_ = [
-    '../../_base_/recog_datasets/ST_MJ_alphanumeric_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
     '../../_base_/schedules/schedule_adam_step_20e.py',
 ]
 
 # dataset settings
-train_list = {{_base_.train_list}}
+train_list = [_base_.mj_rec_train, _base_.st_an_rec_train]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=100))
 
@@ -93,13 +103,25 @@ train_dataloader = dict(
     dataset=dict(
         type='ConcatDataset', datasets=train_list, pipeline=train_pipeline))
 
-visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
+
+visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/crnn/crnn_academic_dataset.py

@@ -1,15 +1,23 @@
 # training schedule for 1x
 _base_ = [
-    'crnn.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
-    '../../_base_/recog_datasets/MJ_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
     '../../_base_/schedules/schedule_adadelta_5e.py',
+    'crnn.py',
 ]
 
 # dataset settings
-train_list = {{_base_.train_list}}
-
+train_list = [_base_.mj_rec_train]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=50), )
 
@@ -53,14 +61,25 @@ train_dataloader = dict(
     sampler=dict(type='DefaultSampler', shuffle=True),
     dataset=dict(
         type='ConcatDataset', datasets=train_list, pipeline=train_pipeline))
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/master/master_r31_12e_ST_MJ_SA.py

@@ -1,13 +1,26 @@
 _base_ = [
-    'master.py',
-    '../../_base_/recog_datasets/ST_SA_MJ_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/synthtext_add.py',
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
     '../../_base_/schedules/schedule_adam_step_12e.py',
+    'master.py',
 ]
 
 # dataset settings
-train_list = {{_base_.train_list}}
+train_list = [
+    _base_.mj_rec_train, _base_.st_rec_train, _base_.st_add_rec_train
+]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=50), )
 
@@ -54,14 +67,25 @@ train_dataloader = dict(
     sampler=dict(type='DefaultSampler', shuffle=True),
     dataset=dict(
         type='ConcatDataset', datasets=train_list, pipeline=train_pipeline))
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/nrtr/nrtr_modality_transform_academic.py

@@ -1,15 +1,26 @@
 _base_ = [
-    'nrtr_modality_transform.py', '../../_base_/recog_datasets/ST_MJ_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
-    '../../_base_/schedules/schedule_adam_step_6e.py'
+    '../../_base_/schedules/schedule_adam_step_6e.py',
+    'nrtr_modality_transform.py',
 ]
 
 # optimizer settings
 optim_wrapper = dict(type='OptimWrapper', optimizer=dict(type='Adam', lr=3e-4))
 
 # dataset settings
-train_list = {{_base_.train_list}}
+train_list = [_base_.mj_rec_train, _base_.st_rec_train]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=50), )
@@ -58,13 +69,25 @@ train_dataloader = dict(
     dataset=dict(
         type='ConcatDataset', datasets=train_list, pipeline=train_pipeline))
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
+
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/nrtr/nrtr_modality_transform_toy_dataset.py

@@ -1,7 +1,8 @@
 _base_ = [
-    'nrtr_modality_transform.py', '../../_base_/default_runtime.py',
     '../../_base_/recog_datasets/toy_data.py',
-    '../../_base_/schedules/schedule_adam_step_6e.py'
+    '../../_base_/default_runtime.py',
+    '../../_base_/schedules/schedule_adam_step_6e.py',
+    'nrtr_modality_transform.py',
 ]
 
 # dataset settings

configs/textrecog/nrtr/nrtr_r31_1by16_1by8_academic.py

@@ -1,15 +1,26 @@
 _base_ = [
-    'nrtr_r31.py', '../../_base_/recog_datasets/ST_MJ_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
-    '../../_base_/schedules/schedule_adam_step_6e.py'
+    '../../_base_/schedules/schedule_adam_step_6e.py',
+    'nrtr_r31.py',
 ]
 
 # optimizer settings
 optim_wrapper = dict(type='OptimWrapper', optimizer=dict(type='Adam', lr=3e-4))
 
 # dataset settings
-train_list = {{_base_.train_list}}
+train_list = [_base_.mj_rec_train, _base_.st_rec_train]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=50))
@@ -58,13 +69,25 @@ train_dataloader = dict(
     dataset=dict(
         type='ConcatDataset', datasets=train_list, pipeline=train_pipeline))
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
+
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/nrtr/nrtr_r31_1by8_1by4_academic.py

@@ -1,15 +1,26 @@
 _base_ = [
-    'nrtr_r31.py', '../../_base_/recog_datasets/ST_MJ_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
-    '../../_base_/schedules/schedule_adam_step_6e.py'
+    '../../_base_/schedules/schedule_adam_step_6e.py',
+    'nrtr_r31.py',
 ]
 
 # optimizer settings
 optim_wrapper = dict(type='OptimWrapper', optimizer=dict(type='Adam', lr=3e-4))
 
 # dataset settings
-train_list = {{_base_.train_list}}
+train_list = [_base_.mj_rec_train, _base_.st_rec_train]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=50), )
@@ -60,13 +71,25 @@ train_dataloader = dict(
     dataset=dict(
         type='ConcatDataset', datasets=train_list, pipeline=train_pipeline))
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
+
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/robust_scanner/robustscanner_r31_academic.py

@@ -1,8 +1,16 @@
 _base_ = [
-    'robust_scanner.py', '../../_base_/recog_datasets/ST_SA_MJ_real_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/coco_text_v1.py'
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
-    '../../_base_/schedules/schedule_adam_step_5e.py'
+    '../../_base_/schedules/schedule_adam_step_5e.py',
+    'robust_scanner.py',
 ]
 
 file_client_args = dict(backend='disk')
@@ -45,47 +53,55 @@ test_pipeline = [
 ]
 
 # dataset settings
-ic11_rec_train = _base_.ic11_rec_train
-ic13_rec_train = _base_.ic13_rec_train
-ic15_rec_train = _base_.ic15_rec_train
-cocov1_rec_train = _base_.cocov1_rec_train
-iiit5k_rec_train = _base_.iiit5k_rec_train
-st_add_rec_train = _base_.st_add_rec_train
-st_rec_train = _base_.st_rec_train
-mj_rec_trian = _base_.mj_rec_trian
+train_list = [
+    _base_.ic11_rec_train, _base_.ic13_rec_train, _base_.ic15_rec_train,
+    _base_.cocov1_rec_train, _base_.iiit5k_rec_train, _base_.mj_sub_rec_train,
+    _base_.st_sub_rec_train, _base_.st_add_rec_train
+]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 
-ic11_rec_train.pipeline = test_pipeline
-ic13_rec_train.pipeline = test_pipeline
-ic15_rec_train.pipeline = test_pipeline
-cocov1_rec_train.pipeline = test_pipeline
-iiit5k_rec_train.pipeline = test_pipeline
-st_add_rec_train.pipeline = test_pipeline
-st_rec_train.pipeline = test_pipeline
-mj_rec_trian.pipeline = test_pipeline
-repeat_ic11 = dict(type='RepeatDataset', dataset=ic11_rec_train, times=20)
-repeat_ic13 = dict(type='RepeatDataset', dataset=ic13_rec_train, times=20)
-repeat_ic15 = dict(type='RepeatDataset', dataset=ic15_rec_train, times=20)
-repeat_cocov1 = dict(type='RepeatDataset', dataset=cocov1_rec_train, times=20)
-repeat_iiit5k = dict(type='RepeatDataset', dataset=iiit5k_rec_train, times=20)
+train_list = [
+    dict(
+        type='RepeatDataset',
+        dataset=dict(
+            type='ConcatDataset',
+            datasets=train_list[:5],
+            pipeline=train_pipeline),
+        times=20),
+    dict(
+        type='ConcatDataset', datasets=train_list[5:],
+        pipeline=train_pipeline),
+]
 
 train_dataloader = dict(
     batch_size=64,
     num_workers=8,
     persistent_workers=True,
     sampler=dict(type='DefaultSampler', shuffle=True),
-    dataset=dict(
-        type='ConcatDataset',
-        datasets=[
-            repeat_ic11, repeat_ic13, repeat_ic15, repeat_cocov1,
-            repeat_iiit5k, st_add_rec_train, st_rec_train, mj_rec_trian
-        ]))
+    dataset=dict(type='ConcatDataset', datasets=train_list))
+
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
+
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/sar/sar_r31_parallel_decoder_academic.py

@@ -1,12 +1,18 @@
 _base_ = [
-    'sar.py',
-    '../../_base_/recog_datasets/ST_SA_MJ_real_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/coco_text_v1.py'
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
     '../../_base_/schedules/schedule_adam_step_5e.py',
+    'sar.py',
 ]
 
-# dataset settings
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=100))
 
@@ -47,47 +53,54 @@ test_pipeline = [
 ]
 
 # dataset settings
-ic11_rec_train = _base_.ic11_rec_train
-ic13_rec_train = _base_.ic13_rec_train
-ic15_rec_train = _base_.ic15_rec_train
-cocov1_rec_train = _base_.cocov1_rec_train
-iiit5k_rec_train = _base_.iiit5k_rec_train
-st_add_rec_train = _base_.st_add_rec_train
-st_rec_train = _base_.st_rec_train
-mj_rec_trian = _base_.mj_rec_trian
+train_list = [
+    _base_.ic11_rec_train, _base_.ic13_rec_train, _base_.ic15_rec_train,
+    _base_.cocov1_rec_train, _base_.iiit5k_rec_train, _base_.mj_sub_rec_train,
+    _base_.st_sub_rec_train, _base_.st_add_rec_train
+]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 
-ic11_rec_train.pipeline = test_pipeline
-ic13_rec_train.pipeline = test_pipeline
-ic15_rec_train.pipeline = test_pipeline
-cocov1_rec_train.pipeline = test_pipeline
-iiit5k_rec_train.pipeline = test_pipeline
-st_add_rec_train.pipeline = test_pipeline
-st_rec_train.pipeline = test_pipeline
-mj_rec_trian.pipeline = test_pipeline
-repeat_ic11 = dict(type='RepeatDataset', dataset=ic11_rec_train, times=20)
-repeat_ic13 = dict(type='RepeatDataset', dataset=ic13_rec_train, times=20)
-repeat_ic15 = dict(type='RepeatDataset', dataset=ic15_rec_train, times=20)
-repeat_cocov1 = dict(type='RepeatDataset', dataset=cocov1_rec_train, times=20)
-repeat_iiit5k = dict(type='RepeatDataset', dataset=iiit5k_rec_train, times=20)
+train_list = [
+    dict(
+        type='RepeatDataset',
+        dataset=dict(
+            type='ConcatDataset',
+            datasets=train_list[:5],
+            pipeline=train_pipeline),
+        times=20),
+    dict(
+        type='ConcatDataset', datasets=train_list[5:],
+        pipeline=train_pipeline),
+]
 
 train_dataloader = dict(
     batch_size=64,
     num_workers=8,
     persistent_workers=True,
     sampler=dict(type='DefaultSampler', shuffle=True),
+    dataset=dict(type='ConcatDataset', datasets=train_list))
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
     dataset=dict(
-        type='ConcatDataset',
-        datasets=[
-            repeat_ic11, repeat_ic13, repeat_ic15, repeat_cocov1,
-            repeat_iiit5k, st_add_rec_train, st_rec_train, mj_rec_trian
-        ]))
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
+
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')

configs/textrecog/sar/sar_r31_sequential_decoder_academic.py

@@ -1,12 +1,18 @@
 _base_ = [
-    'sar.py',
-    '../../_base_/recog_datasets/ST_SA_MJ_real_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/coco_text_v1.py'
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
     '../../_base_/schedules/schedule_adam_step_5e.py',
+    'sar.py',
 ]
 
-# dataset settings
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=100))
 
@@ -47,79 +53,56 @@ test_pipeline = [
 ]
 
 # dataset settings
-ic11_rec_train = _base_.ic11_rec_train
-ic13_rec_train = _base_.ic13_rec_train
-ic15_rec_train = _base_.ic15_rec_train
-cocov1_rec_train = _base_.cocov1_rec_train
-iiit5k_rec_train = _base_.iiit5k_rec_train
-st_add_rec_train = _base_.st_add_rec_train
-st_rec_train = _base_.st_rec_train
-mj_rec_trian = _base_.mj_rec_trian
+train_list = [
+    _base_.ic11_rec_train, _base_.ic13_rec_train, _base_.ic15_rec_train,
+    _base_.cocov1_rec_train, _base_.iiit5k_rec_train, _base_.mj_sub_rec_train,
+    _base_.st_sub_rec_train, _base_.st_add_rec_train
+]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 
-ic11_rec_train.pipeline = train_pipeline
-ic13_rec_train.pipeline = train_pipeline
-ic15_rec_train.pipeline = train_pipeline
-cocov1_rec_train.pipeline = train_pipeline
-iiit5k_rec_train.pipeline = train_pipeline
-st_add_rec_train.pipeline = train_pipeline
-st_rec_train.pipeline = train_pipeline
-mj_rec_trian.pipeline = train_pipeline
-repeat_ic11 = dict(type='RepeatDataset', dataset=ic11_rec_train, times=20)
-repeat_ic13 = dict(type='RepeatDataset', dataset=ic13_rec_train, times=20)
-repeat_ic15 = dict(type='RepeatDataset', dataset=ic15_rec_train, times=20)
-repeat_cocov1 = dict(type='RepeatDataset', dataset=cocov1_rec_train, times=20)
-repeat_iiit5k = dict(type='RepeatDataset', dataset=iiit5k_rec_train, times=20)
+train_list = [
+    dict(
+        type='RepeatDataset',
+        dataset=dict(
+            type='ConcatDataset',
+            datasets=train_list[:5],
+            pipeline=train_pipeline),
+        times=20),
+    dict(
+        type='ConcatDataset', datasets=train_list[5:],
+        pipeline=train_pipeline),
+]
 
 train_dataloader = dict(
     batch_size=64,
     num_workers=16,
     persistent_workers=True,
     sampler=dict(type='DefaultSampler', shuffle=True),
+    dataset=dict(type='ConcatDataset', datasets=train_list))
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
     dataset=dict(
-        type='ConcatDataset',
-        datasets=[
-            repeat_ic11, repeat_ic13, repeat_ic15, repeat_cocov1,
-            repeat_iiit5k, st_add_rec_train, st_rec_train, mj_rec_trian
-        ]))
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
+
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')
 
-dictionary = dict(
-    type='Dictionary',
-    dict_file='dicts/english_digits_symbols.txt',
-    with_start=True,
-    with_end=True,
-    same_start_end=True,
-    with_padding=True,
-    with_unknown=True)
-model = dict(
-    type='SARNet',
-    backbone=dict(type='ResNet31OCR'),
-    encoder=dict(
-        type='SAREncoder',
-        enc_bi_rnn=False,
-        enc_do_rnn=0.1,
-        enc_gru=False,
-    ),
-    decoder=dict(
-        type='SequentialSARDecoder',
-        enc_bi_rnn=False,
-        dec_bi_rnn=False,
-        dec_do_rnn=0,
-        dec_gru=False,
-        pred_dropout=0.1,
-        d_k=512,
-        pred_concat=True,
-        postprocessor=dict(type='AttentionPostprocessor'),
-        module_loss=dict(
-            type='CEModuleLoss', ignore_first_char=True, reduction='mean'),
-        dictionary=dictionary,
-        max_seq_len=30))
+model = dict(decoder=dict(type='SequentialSARDecoder'))

configs/textrecog/satrn/satrn_academic.py

@@ -1,13 +1,23 @@
 _base_ = [
-    'satrn.py',
-    '../../_base_/recog_datasets/ST_MJ_train.py',
-    '../../_base_/recog_datasets/academic_test.py',
+    '../../_base_/recog_datasets/mjsynth.py',
+    '../../_base_/recog_datasets/synthtext.py',
+    '../../_base_/recog_datasets/cute80.py',
+    '../../_base_/recog_datasets/iiit5k.py',
+    '../../_base_/recog_datasets/svt.py',
+    '../../_base_/recog_datasets/svtp.py',
+    '../../_base_/recog_datasets/icdar2013.py',
+    '../../_base_/recog_datasets/icdar2015.py',
     '../../_base_/default_runtime.py',
     '../../_base_/schedules/schedule_adam_step_5e.py',
+    'satrn.py',
 ]
 
 # dataset settings
-train_list = {{_base_.train_list}}
+train_list = [_base_.mj_rec_train, _base_.st_rec_train]
+test_list = [
+    _base_.cute80_rec_test, _base_.iiit5k_rec_test, _base_.svt_rec_test,
+    _base_.svtp_rec_test, _base_.ic13_rec_test, _base_.ic15_rec_test
+]
 file_client_args = dict(backend='disk')
 default_hooks = dict(logger=dict(type='LoggerHook', interval=50))
 
@@ -73,13 +83,25 @@ train_dataloader = dict(
     sampler=dict(type='DefaultSampler', shuffle=True),
     dataset=dict(
         type='ConcatDataset', datasets=train_list, pipeline=train_pipeline))
+test_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type='ConcatDataset', datasets=test_list, pipeline=test_pipeline))
+val_dataloader = test_dataloader
+
+val_evaluator = dict(
+    type='MultiDatasetsEvaluator',
+    metrics=[
+        dict(
+            type='WordMetric',
+            mode=['exact', 'ignore_case', 'ignore_case_symbol']),
+        dict(type='CharMetric')
+    ],
+    datasets_prefix=['CUTE80', 'IIIT5K', 'SVT', 'SVTP', 'IC13', 'IC15'])
+test_evaluator = val_evaluator
 
-test_cfg = dict(type='MultiTestLoop')
-val_cfg = dict(type='MultiValLoop')
-val_dataloader = _base_.val_dataloader
-test_dataloader = _base_.test_dataloader
-for dataloader in test_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
-for dataloader in val_dataloader:
-    dataloader['dataset']['pipeline'] = test_pipeline
 visualizer = dict(type='TextRecogLocalVisualizer', name='visualizer')