[Feature] Support Delving into High-Quality Synthetic Face Occlusion Segmentation Datasets

configs/_base_/datasets/occlude_face.py

@@ -0,0 +1,82 @@
+dataset_type = 'FaceOccludedDataset'
+data_root = 'data/occlusion-aware-face-dataset'
+crop_size = (512, 512)
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='LoadAnnotations'),
+    dict(type='Resize', img_scale=(512, 512)),
+    dict(type='RandomFlip', prob=0.5),
+    dict(type='RandomRotate', degree=(-30, 30), prob=0.5),
+    dict(type='PhotoMetricDistortion'),
+    dict(
+        type='Normalize',
+        mean=[123.675, 116.28, 103.53],
+        std=[58.395, 57.12, 57.375],
+        to_rgb=True),
+    dict(type='DefaultFormatBundle'),
+    dict(type='Collect', keys=['img', 'gt_semantic_seg'])
+]
+test_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(
+        type='MultiScaleFlipAug',
+        img_scale=(512, 512),
+        img_ratios=[0.5, 0.75, 1.0, 1.25, 1.5, 1.75],
+        flip=True,
+        transforms=[
+            dict(type='Resize', keep_ratio=True),
+            dict(type='ResizeToMultiple', size_divisor=32),
+            dict(type='RandomFlip'),
+            dict(
+                type='Normalize',
+                mean=[123.675, 116.28, 103.53],
+                std=[58.395, 57.12, 57.375],
+                to_rgb=True),
+            dict(type='ImageToTensor', keys=['img']),
+            dict(type='Collect', keys=['img'])
+        ])
+]
+
+dataset_train_A = dict(
+    type=dataset_type,
+    data_root=data_root,
+    img_dir='CelebAMask-HQ-original/image',
+    ann_dir='CelebAMask-HQ-original/mask_edited',
+    split='CelebAMask-HQ-original/split/train.txt',
+    pipeline=train_pipeline)
+
+dataset_train_B = dict(
+    type=dataset_type,
+    data_root=data_root,
+    img_dir='NatOcc-SOT/image',
+    ann_dir='NatOcc-SOT/mask',
+    split='NatOcc-SOT/split/train.txt',
+    pipeline=train_pipeline)
+
+
+dataset_valid = dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='RealOcc/image',
+        ann_dir='RealOcc/mask',
+        split='RealOcc/split/val.txt',
+        pipeline=test_pipeline)
+
+dataset_test = dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='RealOcc/image',
+        ann_dir='RealOcc/mask',
+        split='RealOcc/test.txt',
+        pipeline=test_pipeline)
+
+data = dict(
+    samples_per_gpu=2,
+    workers_per_gpu=2,
+    train=[
+            dataset_train_A,dataset_train_B
+    ],
+    val= dataset_valid,
+    test=dataset_test)

configs/deeplabv3plus/deeplabv3plus_r101_512x512_C-CM+C-WO-NatOcc-SOT.py

@@ -0,0 +1,66 @@
+# +
+_base_ = '../_base_/datasets/occlude_face.py'
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    type='EncoderDecoder',
+    pretrained='open-mmlab:https://resnet101_v1c',
+    backbone=dict(
+        type='ResNetV1c',
+        depth=101,
+        num_stages=4,
+        out_indices=(0, 1, 2, 3),
+        dilations=(1, 1, 2, 4),
+        strides=(1, 2, 1, 1),
+        norm_cfg=dict(type='SyncBN', requires_grad=True),
+        norm_eval=False,
+        style='pytorch',
+        contract_dilation=True),
+    decode_head=dict(
+        type='DepthwiseSeparableASPPHead',
+        in_channels=2048,
+        in_index=3,
+        channels=512,
+        dilations=(1, 12, 24, 36),
+        c1_in_channels=256,
+        c1_channels=48,
+        dropout_ratio=0.1,
+        num_classes=2,
+        norm_cfg=dict(type='SyncBN', requires_grad=True),
+        align_corners=False,
+        loss_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000)),
+    auxiliary_head=dict(
+        type='FCNHead',
+        in_channels=1024,
+        in_index=2,
+        channels=256,
+        num_convs=1,
+        concat_input=False,
+        dropout_ratio=0.1,
+        num_classes=2,
+        norm_cfg=dict(type='SyncBN', requires_grad=True),
+        align_corners=False,
+        loss_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
+    train_cfg=dict(),
+    test_cfg=dict(mode='whole'))
+log_config = dict(
+    interval=50, hooks=[dict(type='TextLoggerHook', by_epoch=False)])
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
+cudnn_benchmark = True
+optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0005)
+optimizer_config = dict()
+lr_config = dict(policy='poly', power=0.9, min_lr=0.0001, by_epoch=False)
+runner = dict(type='IterBasedRunner', max_iters=30000)
+checkpoint_config = dict(by_epoch=False, interval=400)
+evaluation = dict(
+    interval=400, metric=['mIoU', 'mDice', 'mFscore'], pre_eval=True)
+
+work_dir = './work_dirs/deeplabv3plus_r101_512x512_C-CM+C-WO-NatOcc-SOT'
+gpu_ids = range(0, 2)
+auto_resume = False

docs/en/dataset_prepare.md

@@ -1,3 +1,4 @@
+<!-- #region -->
 ## Prepare datasets
 
 It is recommended to symlink the dataset root to `$MMSEGMENTATION/data`.
@@ -376,3 +377,58 @@ python tools/convert_datasets/isaid.py /path/to/iSAID
 ```
 
 In our default setting (`patch_width`=896, `patch_height`=896,　`overlap_area`=384), it will generate 33978 images for training and 11644 images for validation.
+
+
+### Delving into High-Quality Synthetic Face Occlusion Segmentation Datasets
+
+The dataset is generated by two techniques, Naturalistic occlusion generation, Random occlusion generation. you must install face-occlusion-generation and dataset. see more guide in https://github.com/kennyvoo/face-occlusion-generation.git
+
+
+## Dataset Preparation
+
+Please download the masks from this [drive](https://drive.google.com/drive/folders/15nZETWlGMdcKY6aHbchRsWkUI42KTNs5?usp=sharing) and the images from [CelebAMask-HQ](https://github.com/switchablenorms/CelebAMask-HQ), [11k Hands](https://sites.google.com/view/11khands) and [DTD](https://www.robots.ox.ac.uk/~vgg/data/dtd/). 
+
+The extracted and upsampled COCO objects images and masks can be found in this [drive](https://drive.google.com/drive/folders/15nZETWlGMdcKY6aHbchRsWkUI42KTNs5?usp=sharing).
+
+Please extract CelebAMask-HQ and 11k Hands images based on the splits found in [drive](https://drive.google.com/drive/folders/15nZETWlGMdcKY6aHbchRsWkUI42KTNs5?usp=sharing). 
+
+**Dataset Organization:**
+
+```none
+
+├── dataset
+│   ├── CelebAMask-HQ-WO-Train_img
+│   │   ├── {image}.jpg
+│   ├── CelebAMask-HQ-WO-Train_mask
+│   │   ├── {mask}.png
+│   ├── DTD
+│   │   ├── images
+│   │   │   ├── {classA}
+│   │   │   │   ├── {image}.jpg
+│   │   │   ├── {classB}
+│   │   │   │   ├── {image}.jpg
+│   ├── 11k-hands_img
+│   │   ├── {image}.jpg
+│   ├── 11k-hands_mask
+│   │   ├── {mask}.png
+│   ├── object_image_sr
+│   │   ├── {image}.jpg
+│   ├── object_image_x4
+│   │   ├── {mask}.png
+
+```
+
+## Data Generation
+
+Example script to generate NatOcc dataset 
+
+bash NatOcc.sh
+
+Example script to generate RandOcc dataset
+
+bash RandOcc.sh
+<!-- #endregion -->
+
+```python
+
+```

mmseg/datasets/__init__.py

@@ -18,6 +18,7 @@
 from .potsdam import PotsdamDataset
 from .stare import STAREDataset
 from .voc import PascalVOCDataset
+from .face import FaceOccludedDataset
 
 __all__ = [
     'CustomDataset', 'build_dataloader', 'ConcatDataset', 'RepeatDataset',
@@ -26,5 +27,4 @@
     'PascalContextDataset59', 'ChaseDB1Dataset', 'DRIVEDataset', 'HRFDataset',
     'STAREDataset', 'DarkZurichDataset', 'NightDrivingDataset',
     'COCOStuffDataset', 'LoveDADataset', 'MultiImageMixDataset',
-    'iSAIDDataset', 'ISPRSDataset', 'PotsdamDataset'
-]
+    'iSAIDDataset', 'ISPRSDataset', 'PotsdamDataset', 'FaceOccludedDataset']

mmseg/datasets/face.py

@@ -0,0 +1,23 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import os.path as osp
+
+from .builder import DATASETS
+from .custom import CustomDataset
+
+
+@DATASETS.register_module()
+class FaceOccludedDataset(CustomDataset):
+    """Face Occluded dataset.
+
+    Args:
+        split (str): Split txt file for Pascal VOC.
+    """
+
+    CLASSES = ('background', 'face')
+
+    PALETTE = [[0, 0, 0], [128, 0, 0]]
+
+    def __init__(self, split, **kwargs):
+        super(FaceOccludedDataset, self).__init__(
+            img_suffix='.jpg', seg_map_suffix='.png', split=split, **kwargs)
+        assert osp.exists(self.img_dir) and self.split is not None