support coco stuff-10k/164k

configs/_base_/datasets/coco_stuff10k.py

@@ -0,0 +1,57 @@
+# dataset settings
+dataset_type = 'COCOStuffDataset'
+data_root = 'data/coco_stuff10k'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+crop_size = (512, 512)
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='LoadAnnotations', reduce_zero_label=True),
+    dict(type='Resize', img_scale=(2048, 512), 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=(2048, 512),
+        # img_ratios=[0.5, 0.75, 1.0, 1.25, 1.5, 1.75],
+        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=dataset_type,
+        data_root=data_root,
+        reduce_zero_label=True,
+        img_dir='images/train2014',
+        ann_dir='masks/train2014',
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        data_root=data_root,
+        reduce_zero_label=True,
+        img_dir='images/test2014',
+        ann_dir='masks/test2014',
+        pipeline=test_pipeline),
+    test=dict(
+        type=dataset_type,
+        data_root=data_root,
+        reduce_zero_label=True,
+        img_dir='images/test2014',
+        ann_dir='masks/test2014',
+        pipeline=test_pipeline))

configs/_base_/datasets/coco_stuff164k.py

@@ -0,0 +1,54 @@
+# dataset settings
+dataset_type = 'COCOStuffDataset'
+data_root = 'data/coco_stuff164k'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+crop_size = (512, 512)
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='LoadAnnotations'),
+    dict(type='Resize', img_scale=(2048, 512), 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=(2048, 512),
+        # img_ratios=[0.5, 0.75, 1.0, 1.25, 1.5, 1.75],
+        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=dataset_type,
+        data_root=data_root,
+        img_dir='images/train2017',
+        ann_dir='masks/train2017',
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='images/val2017',
+        ann_dir='masks/val2017',
+        pipeline=test_pipeline),
+    test=dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='images/val2017',
+        ann_dir='masks/val2017',
+        pipeline=test_pipeline))

configs/pspnet/pspnet_r101-d8_512x512_80k_coco-stuff10k.py

@@ -0,0 +1,2 @@
+_base_ = './pspnet_r50-d8_512x512_80k_coco-stuff10k.py'
+model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))

configs/pspnet/pspnet_r101-d8_512x512_80k_coco-stuff164k.py

@@ -0,0 +1,2 @@
+_base_ = './pspnet_r50-d8_512x512_80k_coco-stuff164k.py'
+model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))

configs/pspnet/pspnet_r50-d8_512x512_80k_coco-stuff10k.py

@@ -0,0 +1,6 @@
+_base_ = [
+    '../_base_/models/pspnet_r50-d8.py', '../_base_/datasets/coco_stuff10k.py',
+    '../_base_/default_runtime.py', '../_base_/schedules/schedule_80k.py'
+]
+model = dict(
+    decode_head=dict(num_classes=171), auxiliary_head=dict(num_classes=171))

configs/pspnet/pspnet_r50-d8_512x512_80k_coco-stuff164k.py

@@ -0,0 +1,8 @@
+_base_ = [
+    '../_base_/models/pspnet_r50-d8.py',
+    '../_base_/datasets/coco_stuff164k.py',
+    '../_base_/default_runtime.py',
+    '../_base_/schedules/schedule_80k.py'
+]
+model = dict(
+    decode_head=dict(num_classes=171), auxiliary_head=dict(num_classes=171))

docs/dataset_prepare.md

@@ -41,6 +41,28 @@ mmsegmentation
 │   │   │   ├── images
 │   │   │   │   ├── training
 │   │   │   │   ├── validation
+│   ├── coco_stuff10k
+│   │   ├── images
+│   │   │   ├── train2014
+│   │   │   ├── test2014
+│   │   ├── masks
+│   │   │   ├── train2014
+│   │   │   ├── test2014
+│   │   ├── annotations
+│   │   ├── imagesLists
+│   │   │   ├── train.txt
+│   │   │   ├── test.txt
+│   │   │   ├── all.txt
+│   ├── coco_stuff164k
+│   │   ├── images
+│   │   │   ├── train2017
+│   │   │   ├── val2017
+│   │   ├── masks
+│   │   │   ├── train2017
+│   │   │   ├── val2017
+│   │   ├── annotations
+│   │   │   ├── train2017
+│   │   │   ├── val2017
 │   ├── CHASE_DB1
 │   │   ├── images
 │   │   │   ├── training
@@ -85,6 +107,54 @@ to generate `**labelTrainIds.png`.
 python tools/convert_datasets/cityscapes.py data/cityscapes --nproc 8
 ```
 
+### COCO Stuff 10k
+
+The data could be downloaded [here](http://calvin.inf.ed.ac.uk/wp-content/uploads/data/cocostuffdataset/cocostuff-10k-v1.1.zip) by wget.
+
+If you would like to use COCO Stuff 10k dataset, please run following command to convert augmentation annotations into proper format.
+
+```shell
+# --nproc means 8 process for conversion, which could be omitted as well.
+python tools/convert_datasets/coco_stuff10k.py data/coco_stuff10k --nproc 8
+```
+
+By convention, masks in `data/coco_stuff10k/masks/*.png` are used for COCO Stuff 10k training and testing.
+
+
+### COCO Stuff 164k
+
+The data could be downloaded and converted augmentation annotations into proper format. by following command.
+
+```shell
+cd data
+
+# create file
+mkdir -p coco_stuff164k
+cd coco_stuff164k
+
+# Download everything
+wget --directory-prefix=downloads http://images.cocodataset.org/zips/train2017.zip
+wget --directory-prefix=downloads http://images.cocodataset.org/zips/val2017.zip
+wget --directory-prefix=downloads http://calvin.inf.ed.ac.uk/wp-content/uploads/data/cocostuffdataset/stuffthingmaps_trainval2017.zip
+
+# Unpack everything
+mkdir -p images
+mkdir -p annotations
+unzip downloads/train2017.zip -d images/
+unzip downloads/val2017.zip -d images/
+unzip downloads/stuffthingmaps_trainval2017.zip -d annotations/
+
+cd ../..
+
+# --nproc means 8 process for conversion, which could be omitted as well.
+python tools/convert_datasets/coco_stuff164k.py data/coco_stuff164k --nproc 8
+```
+
+By convention, masks in `data/coco_stuff164k/masks/*.png` are used for COCO Stuff 164k training and testing.
+
+The details of this dataset could be found at [here](https://github.com/nightrome/cocostuff#downloads).
+
+
 ### Pascal VOC
 
 Pascal VOC 2012 could be downloaded from [here](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar).

mmseg/datasets/__init__.py

@@ -9,11 +9,12 @@
 from .pascal_context import PascalContextDataset, PascalContextDataset59
 from .stare import STAREDataset
 from .voc import PascalVOCDataset
+from .coco_stuff import COCOStuffDataset
 
 __all__ = [
     'CustomDataset', 'build_dataloader', 'ConcatDataset', 'RepeatDataset',
     'DATASETS', 'build_dataset', 'PIPELINES', 'CityscapesDataset',
     'PascalVOCDataset', 'ADE20KDataset', 'PascalContextDataset',
     'PascalContextDataset59', 'ChaseDB1Dataset', 'DRIVEDataset', 'HRFDataset',
-    'STAREDataset'
+    'STAREDataset', 'COCOStuffDataset'
 ]