[DBNet] Add DBNet config

configs/_base_/det_models/dbnet_r18_fpnc.py

@@ -1,3 +1,9 @@
+preprocess_cfg = dict(
+    mean=[123.675, 116.28, 103.53],
+    std=[58.395, 57.12, 57.375],
+    to_rgb=True,
+    pad_size_divisor=32)
+
 model = dict(
     type='DBNet',
     backbone=dict(
@@ -12,10 +18,9 @@ model = dict(
         style='caffe'),
     neck=dict(
         type='FPNC', in_channels=[64, 128, 256, 512], lateral_channels=256),
-    bbox_head=dict(
+    det_head=dict(
         type='DBHead',
         in_channels=256,
-        loss=dict(type='DBLoss', alpha=5.0, beta=10.0, bbce_loss=True),
+        loss=dict(type='DBLoss'),
         postprocessor=dict(type='DBPostprocessor', text_repr_type='quad')),
-    train_cfg=None,
+    preprocess_cfg=preprocess_cfg)
-    test_cfg=None)

configs/_base_/det_models/dbnet_r50dcnv2_fpnc.py

@@ -1,3 +1,9 @@
+preprocess_cfg = dict(
+    mean=[122.67891434, 116.66876762, 104.00698793],
+    std=[58.395, 57.12, 57.375],
+    to_rgb=True,
+    pad_size_divisor=32)
+
 model = dict(
     type='DBNet',
     backbone=dict(
@@ -14,10 +20,9 @@ model = dict(
         stage_with_dcn=(False, True, True, True)),
     neck=dict(
         type='FPNC', in_channels=[256, 512, 1024, 2048], lateral_channels=256),
-    bbox_head=dict(
+    det_head=dict(
         type='DBHead',
         in_channels=256,
-        loss=dict(type='DBLoss', alpha=5.0, beta=10.0, bbce_loss=True),
+        loss=dict(type='DBLoss'),
         postprocessor=dict(type='DBPostprocessor', text_repr_type='quad')),
-    train_cfg=None,
+    preprocess_cfg=preprocess_cfg)
-    test_cfg=None)

configs/textdet/dbnet/dbnet_r18_fpnc_1200e_icdar2015.py

@@ -2,32 +2,78 @@ _base_ = [
     '../../_base_/default_runtime.py',
     '../../_base_/schedules/schedule_sgd_1200e.py',
     '../../_base_/det_models/dbnet_r18_fpnc.py',
-    '../../_base_/det_datasets/icdar2015.py',
-    '../../_base_/det_pipelines/dbnet_pipeline.py'
 ]
 
-train_list = {{_base_.train_list}}
+default_hooks = dict(checkpoint=dict(type='CheckpointHook', interval=20), )
-test_list = {{_base_.test_list}}
 
-train_pipeline_r18 = {{_base_.train_pipeline_r18}}
+train_pipeline_r18 = [
-test_pipeline_1333_736 = {{_base_.test_pipeline_1333_736}}
+    dict(type='LoadImageFromFile', color_type='color_ignore_orientation'),
+    dict(
+        type='LoadOCRAnnotations',
+        with_polygon=True,
+        with_bbox=True,
+        with_label=True,
+    ),
+    dict(
+        type='TorchVisionWrapper',
+        op='ColorJitter',
+        brightness=32.0 / 255,
+        saturation=0.5),
+    dict(
+        type='ImgAug',
+        args=[['Fliplr', 0.5],
+              dict(cls='Affine', rotate=[-10, 10]), ['Resize', [0.5, 3.0]]]),
+    dict(type='RandomCrop', min_side_ratio=0.1),
+    dict(type='Resize', scale=(640, 640), keep_ratio=True),
+    dict(type='Pad', size=(640, 640)),
+    dict(
+        type='PackTextDetInputs',
+        meta_keys=('img_path', 'ori_shape', 'img_shape'))
+]
 
-data = dict(
+test_pipeline_1333_736 = [
-    samples_per_gpu=16,
+    dict(type='LoadImageFromFile', color_type='color_ignore_orientation'),
-    workers_per_gpu=8,
+    dict(type='Resize', scale=(1333, 736), keep_ratio=True),
-    val_dataloader=dict(samples_per_gpu=1),
+    dict(
-    test_dataloader=dict(samples_per_gpu=1),
+        type='PackTextDetInputs',
-    train=dict(
+        meta_keys=('img_path', 'ori_shape', 'img_shape', 'scale_factor',
-        type='UniformConcatDataset',
+                   'instances'))
-        datasets=train_list,
+]
-        pipeline=train_pipeline_r18),
-    val=dict(
-        type='UniformConcatDataset',
-        datasets=test_list,
-        pipeline=test_pipeline_1333_736),
-    test=dict(
-        type='UniformConcatDataset',
-        datasets=test_list,
-        pipeline=test_pipeline_1333_736))
 
-evaluation = dict(interval=100, metric='hmean-iou')
+dataset_type = 'OCRDataset'
+data_root = 'data/icdar2015'
+
+train_dataset = dict(
+    type=dataset_type,
+    data_root=data_root,
+    ann_file='instances_training.json',
+    data_prefix=dict(img_path='imgs/'),
+    filter_cfg=dict(filter_empty_gt=True, min_size=32),
+    pipeline=train_pipeline_r18)
+
+test_dataset = dict(
+    type=dataset_type,
+    data_root=data_root,
+    ann_file='instances_test.json',
+    data_prefix=dict(img_path='imgs/'),
+    test_mode=True,
+    pipeline=test_pipeline_1333_736)
+
+train_dataloader = dict(
+    batch_size=16,
+    num_workers=8,
+    persistent_workers=False,
+    sampler=dict(type='DefaultSampler', shuffle=True),
+    dataset=train_dataset)
+val_dataloader = dict(
+    batch_size=16,
+    num_workers=8,
+    persistent_workers=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=test_dataset)
+test_dataloader = val_dataloader
+
+val_evaluator = dict(type='HmeanIOUMetric')
+test_evaluator = val_evaluator
+
+visualizer = dict(type='TextDetLocalVisualizer', name='visualizer')

configs/textdet/dbnet/dbnet_r50dcnv2_fpnc_1200e_icdar2015.py

@@ -2,34 +2,80 @@ _base_ = [
     '../../_base_/default_runtime.py',
     '../../_base_/schedules/schedule_sgd_1200e.py',
     '../../_base_/det_models/dbnet_r50dcnv2_fpnc.py',
-    '../../_base_/det_datasets/icdar2015.py',
-    '../../_base_/det_pipelines/dbnet_pipeline.py'
 ]
 
-train_list = {{_base_.train_list}}
+default_hooks = dict(checkpoint=dict(type='CheckpointHook', interval=20), )
-test_list = {{_base_.test_list}}
-
-train_pipeline_r50dcnv2 = {{_base_.train_pipeline_r50dcnv2}}
-test_pipeline_4068_1024 = {{_base_.test_pipeline_4068_1024}}
 
 load_from = 'checkpoints/textdet/dbnet/res50dcnv2_synthtext.pth'
 
-data = dict(
+train_pipeline_r50dcnv2 = [
-    samples_per_gpu=8,
+    dict(type='LoadImageFromFile', color_type='color_ignore_orientation'),
-    workers_per_gpu=4,
+    dict(
-    val_dataloader=dict(samples_per_gpu=1),
+        type='LoadOCRAnnotations',
-    test_dataloader=dict(samples_per_gpu=1),
+        with_bbox=True,
-    train=dict(
+        with_polygon=True,
-        type='UniformConcatDataset',
+        with_label=True,
-        datasets=train_list,
+    ),
-        pipeline=train_pipeline_r50dcnv2),
+    dict(
-    val=dict(
+        type='TorchVisionWrapper',
-        type='UniformConcatDataset',
+        op='ColorJitter',
-        datasets=test_list,
+        brightness=32.0 / 255,
-        pipeline=test_pipeline_4068_1024),
+        saturation=0.5),
-    test=dict(
+    dict(
-        type='UniformConcatDataset',
+        type='ImgAug',
-        datasets=test_list,
+        args=[['Fliplr', 0.5],
-        pipeline=test_pipeline_4068_1024))
+              dict(cls='Affine', rotate=[-10, 10]), ['Resize', [0.5, 3.0]]]),
+    dict(type='RandomCrop', min_side_ratio=0.1),
+    dict(type='Resize', scale=(640, 640), keep_ratio=True),
+    dict(type='Pad', size=(640, 640)),
+    dict(
+        type='PackTextDetInputs',
+        meta_keys=('img_path', 'ori_shape', 'img_shape'))
+]
 
-evaluation = dict(interval=100, metric='hmean-iou')
+test_pipeline_4068_1024 = [
+    dict(type='LoadImageFromFile', color_type='color_ignore_orientation'),
+    dict(type='Resize', scale=(4068, 1024), keep_ratio=True),
+    dict(
+        type='PackTextDetInputs',
+        meta_keys=('img_path', 'ori_shape', 'img_shape', 'scale_factor',
+                   'instances'))
+]
+
+dataset_type = 'OCRDataset'
+data_root = 'data/icdar2015'
+
+train_dataset = dict(
+    type=dataset_type,
+    data_root=data_root,
+    ann_file='instances_training.json',
+    data_prefix=dict(img_path='imgs/'),
+    filter_cfg=dict(filter_empty_gt=True, min_size=32),
+    pipeline=train_pipeline_r50dcnv2)
+
+test_dataset = dict(
+    type=dataset_type,
+    data_root=data_root,
+    ann_file='instances_test.json',
+    data_prefix=dict(img_path='imgs/'),
+    test_mode=True,
+    pipeline=test_pipeline_4068_1024)
+
+train_dataloader = dict(
+    batch_size=16,
+    num_workers=8,
+    persistent_workers=True,
+    sampler=dict(type='DefaultSampler', shuffle=True),
+    dataset=train_dataset)
+val_dataloader = dict(
+    batch_size=16,
+    num_workers=8,
+    persistent_workers=True,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=test_dataset)
+test_dataloader = val_dataloader
+
+val_evaluator = dict(type='HmeanIOUMetric')
+test_evaluator = val_evaluator
+
+visualizer = dict(type='TextDetLocalVisualizer', name='visualizer')