Custom Bisenetv2 config for new dataset

configs/_base_/datasets/waterpuddles.py

@@ -0,0 +1,59 @@
+# dataset settings
+dataset_type = 'waterpuddlesDataset'
+data_root = 'data/puddle-1000_chasedb_NEW'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=False)
+img_scale = (360, 640)
+crop_size = (128, 128)
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='LoadAnnotations'),
+    dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
+    dict(type='RandomFlip', prob=0.5),
+    dict(type='PhotoMetricDistortion'),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+    dict(type='DefaultFormatBundle'),
+    dict(type='Collect', keys=['img', 'gt_semantic_seg'])
+]
+test_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(
+        type='MultiScaleFlipAug',
+        img_scale=img_scale,
+        # img_ratios=[0.5, 0.75, 1.0, 1.25, 1.5, 1.75, 2.0],
+        flip=False,
+        transforms=[
+            dict(type='Resize', keep_ratio=True),
+            dict(type='RandomFlip'),
+            dict(type='Normalize', **img_norm_cfg),
+            dict(type='ImageToTensor', keys=['img']),
+            dict(type='Collect', keys=['img'])
+        ])
+]
+
+data = dict(
+    samples_per_gpu=4,
+    workers_per_gpu=4,
+    train=dict(
+        type='RepeatDataset',
+        times=40000,
+        dataset=dict(
+            type=dataset_type,
+            data_root=data_root,
+            img_dir='images/training',
+            ann_dir='annotations/training',
+            pipeline=train_pipeline)),
+    val=dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='images/validation',
+        ann_dir='annotations/validation',
+        pipeline=test_pipeline),
+    test=dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='images/validation',
+        ann_dir='annotations/validation',
+        pipeline=test_pipeline))

configs/_base_/models/bisenetv2.py

@@ -20,7 +20,7 @@
         num_convs=1,
         concat_input=False,
         dropout_ratio=0.1,
-        num_classes=19,
+        num_classes=2,
         norm_cfg=norm_cfg,
         align_corners=False,
         loss_decode=dict(
@@ -31,7 +31,7 @@
             in_channels=16,
             channels=16,
             num_convs=2,
-            num_classes=19,
+            num_classes=2,
             in_index=1,
             norm_cfg=norm_cfg,
             concat_input=False,
@@ -43,7 +43,7 @@
             in_channels=32,
             channels=64,
             num_convs=2,
-            num_classes=19,
+            num_classes=2,
             in_index=2,
             norm_cfg=norm_cfg,
             concat_input=False,
@@ -55,7 +55,7 @@
             in_channels=64,
             channels=256,
             num_convs=2,
-            num_classes=19,
+            num_classes=2,
             in_index=3,
             norm_cfg=norm_cfg,
             concat_input=False,
@@ -67,7 +67,7 @@
             in_channels=128,
             channels=1024,
             num_convs=2,
-            num_classes=19,
+            num_classes=2,
             in_index=4,
             norm_cfg=norm_cfg,
             concat_input=False,

configs/_base_/schedules/schedule_40k.py

@@ -5,5 +5,5 @@
 lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
 # runtime settings
 runner = dict(type='IterBasedRunner', max_iters=40000)
-checkpoint_config = dict(by_epoch=False, interval=4000)
-evaluation = dict(interval=4000, metric='mIoU', pre_eval=True)
+checkpoint_config = dict(by_epoch=False, interval=2000)
+evaluation = dict(interval=2000, metric='mIoU', pre_eval=True)

configs/bisenetv2/bisenetv2_fcn_4x4_512x512_40k_waterpuddles.py

@@ -0,0 +1,6 @@
+_base_ = [
+    '../_base_/models/bisenetv2.py',    # model settings
+    '../_base_/datasets/waterpuddles.py', # dataset settings
+    '../_base_/schedules/schedule_40k.py', # scheduler settings
+    '../_base_/default_runtime.py' # other runtime settings
+]

mmseg/core/evaluation/class_names.py

@@ -120,10 +120,11 @@ def isaid_classes():
         'Soccer_ball_field', 'plane', 'Harbor'
     ]
 
-
-def stare_classes():
-    """stare class names for external use."""
-    return ['background', 'vessel']
+def waterpuddles_classes():
+    """waterpuddles class names for external use."""
+    return [
+        'background', 'waterpuddle'
+    ]
 
 
 def cityscapes_palette():
@@ -259,9 +260,8 @@ def isaid_palette():
             [0, 0, 191], [0, 0, 255], [0, 191, 127], [0, 127, 191],
             [0, 127, 255], [0, 100, 155]]
 
-
-def stare_palette():
-    """STARE palette for external use."""
+def waterpuddles_palette():
+    """waterpuddles palette for external use."""
     return [[120, 120, 120], [6, 230, 230]]
 
 
@@ -278,7 +278,7 @@ def stare_palette():
         'coco_stuff164k'
     ],
     'isaid': ['isaid', 'iSAID'],
-    'stare': ['stare', 'STARE']
+    'waterpuddles': ['waterpuddles', 'waterpuddle']
 }
 
 

mmseg/datasets/waterpuddles.py

@@ -0,0 +1,27 @@
+import os.path as osp
+
+from .builder import DATASETS
+from .custom import CustomDataset
+
+
+@DATASETS.register_module()
+class waterpuddlesDataset(CustomDataset):
+    """water puddles dataset.
+
+    In segmentation map annotation for Chase_db1, 0 stands for background,
+    which is included in 2 categories. ``reduce_zero_label`` is fixed to False.
+    The ``img_suffix`` is fixed to '_img.png' and ``seg_map_suffix`` is fixed to
+    '_anno.png'.
+    """
+
+    CLASSES = ('background','waterpuddle')
+
+    PALETTE = [[120, 120, 120], [6, 230, 230]]
+
+    def __init__(self, **kwargs):
+        super(waterpuddlesDataset, self).__init__(
+            img_suffix='.png',
+            seg_map_suffix='.png',
+            reduce_zero_label=False,
+            **kwargs)
+        assert osp.exists(self.img_dir)