resolve conflicts

.github/workflows/merge_stage_test.yml

@@ -37,9 +37,6 @@ jobs:
         run: |
           pip install pip --upgrade
           pip install wheel
-      - name: Install Pillow
-        run: pip install Pillow==6.2.2
-        if: ${{matrix.torchvision == '0.4.2'}}
       - name: Install PyTorch
         run: pip install torch==${{matrix.torch}}+cpu torchvision==${{matrix.torchvision}}+cpu -f https://download.pytorch.org/whl/torch_stable.html
       - name: Install mmseg dependencies
@@ -65,7 +62,8 @@ jobs:
     strategy:
       matrix:
         python-version: [3.7]
-        torch: [1.6.0, 1.7.1, 1.8.1, 1.9.1, 1.10.1, 1.11.0, 1.12.1, 1.13.0]
+        torch: [1.6.0, 1.7.1, 1.8.1, 1.9.1, 1.10.1, 1.11.0, 1.12.0]
+        # torch: [1.6.0, 1.7.1, 1.8.1, 1.9.1, 1.10.1, 1.11.0, 1.12.0, 1.13.0]
         include:
           - torch: 1.6.0
             torchvision: 0.7.0
@@ -79,10 +77,10 @@ jobs:
             torchvision: 0.11.2
           - torch: 1.11.0
             torchvision: 0.12.0
-          - torch: 1.12.1
+          - torch: 1.12.0
             torchvision: 0.13.0
-          - torch: 1.13.0
-            torchvision:
+          # - torch: 1.13.0
+          #   torchvision: 0.14.0
     steps:
       - uses: actions/checkout@v2
       - name: Set up Python ${{ matrix.python-version }}
@@ -160,7 +158,6 @@ jobs:
           apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/7fa2af80.pub
       - name: Install Python-dev
         run: apt-get update && apt-get install -y python${{matrix.python-version}}-dev
-        if: ${{matrix.python-version != 3.9}}
       - name: Install system dependencies
         run: |
           apt-get update && apt-get install -y ffmpeg libsm6 libxext6 git ninja-build libglib2.0-0 libsm6 libxrender-dev libxext6

docs/zh_cn/advanced_guides/datasets.md

@@ -1,4 +1,4 @@
-#　数据集
+# 数据集
 
 在 MMSegmentation 算法库中, 所有 Dataset 类的功能有两个: 加载[预处理](https://github.com/open-mmlab/mmsegmentation/blob/dev-1.x/docs/en/user_guides/2_dataset_prepare.md) 之后的数据集的信息, 和将数据送入[数据集变换流水线](https://github.com/open-mmlab/mmsegmentation/blob/dev-1.x/mmseg/datasets/basesegdataset.py#L141) 中, 进行[数据变换操作](https://github.com/open-mmlab/mmsegmentation/blob/dev-1.x/docs/zh_cn/advanced_guides/transforms.md). 加载的数据集信息包括两类: 元信息 (meta information), 数据集本身的信息, 例如数据集总共的类别, 和它们对应调色盘信息: 数据信息 (data information) 是指每组数据中图片和对应标签的路径. 下文中介绍了 MMSegmentation 1.x 中数据集的常用接口, 和 mmseg 数据集基类中数据信息加载与修改数据集类别的逻辑, 以及数据集与数据变换流水线 (pipeline) 的关系.
 

mmseg/datasets/__init__.py

@@ -10,7 +10,6 @@
 from .decathlon import DecathlonDataset
 from .drive import DRIVEDataset
 from .hrf import HRFDataset
-from .refuge import REFUGEDataset
 from .isaid import iSAIDDataset
 from .isprs import ISPRSDataset
 from .lip import LIPDataset
@@ -28,20 +27,23 @@
                          PackSegInputs, PhotoMetricDistortion, RandomCrop,
                          RandomCutOut, RandomMosaic, RandomRotate, Rerange,
                          ResizeShortestEdge, ResizeToMultiple, RGB2Gray,
-                         SegRescale,Rot90,ColorJitter,RandomFlipud,RandomFliplr)
+                         SegRescale)
 from .voc import PascalVOCDataset
 
 # yapf: enable
 __all__ = [
-    'BaseSegDataset', 'CityscapesDataset', 'PascalVOCDataset', 'ADE20KDataset',
-    'PascalContextDataset', 'PascalContextDataset59', 'ChaseDB1Dataset',
-    'DRIVEDataset', 'HRFDataset', 'STAREDataset', 'DarkZurichDataset',
-    'NightDrivingDataset', 'COCOStuffDataset', 'LoveDADataset',
-    'MultiImageMixDataset', 'iSAIDDataset', 'ISPRSDataset', 'PotsdamDataset',
-    'LoadAnnotations', 'RandomCrop', 'SegRescale', 'PhotoMetricDistortion',
-    'RandomRotate', 'AdjustGamma', 'CLAHE', 'Rerange', 'RGB2Gray','ColorJitter',
-    'RandomCutOut', 'RandomMosaic', 'PackSegInputs', 'ResizeToMultiple',
-    'LoadImageFromNDArray', 'LoadBiomedicalImageFromFile','Rot90','RandomFlipud','RandomFliplr',
+    'BaseSegDataset', 'BioMedical3DRandomCrop', 'CityscapesDataset',
+    'PascalVOCDataset', 'ADE20KDataset', 'PascalContextDataset',
+    'PascalContextDataset59', 'ChaseDB1Dataset', 'DRIVEDataset', 'HRFDataset',
+    'STAREDataset', 'DarkZurichDataset', 'NightDrivingDataset',
+    'COCOStuffDataset', 'LoveDADataset', 'MultiImageMixDataset',
+    'iSAIDDataset', 'ISPRSDataset', 'PotsdamDataset', 'LoadAnnotations',
+    'RandomCrop', 'SegRescale', 'PhotoMetricDistortion', 'RandomRotate',
+    'AdjustGamma', 'CLAHE', 'Rerange', 'RGB2Gray', 'RandomCutOut',
+    'RandomMosaic', 'PackSegInputs', 'ResizeToMultiple',
+    'LoadImageFromNDArray', 'LoadBiomedicalImageFromFile',
     'LoadBiomedicalAnnotation', 'LoadBiomedicalData', 'GenerateEdge',
-    'DecathlonDataset', 'LIPDataset', 'ResizeShortestEdge','REFUGEDataset'
+    'DecathlonDataset', 'LIPDataset', 'ResizeShortestEdge',
+    'BioMedicalGaussianNoise', 'BioMedicalGaussianBlur',
+    'BioMedicalRandomGamma'
 ]

mmseg/datasets/transforms/__init__.py

@@ -10,7 +10,7 @@
                          PhotoMetricDistortion, RandomCrop, RandomCutOut,
                          RandomMosaic, RandomRotate, Rerange,
                          ResizeShortestEdge, ResizeToMultiple, RGB2Gray,
-                         SegRescale,Rot90,ColorJitter,RandomFlipud,RandomFliplr)
+                         SegRescale)
 
 # yapf: enable
 __all__ = [
@@ -19,5 +19,6 @@
     'RGB2Gray', 'RandomCutOut', 'RandomMosaic', 'PackSegInputs',
     'ResizeToMultiple', 'LoadImageFromNDArray', 'LoadBiomedicalImageFromFile',
     'LoadBiomedicalAnnotation', 'LoadBiomedicalData', 'GenerateEdge',
-    'ResizeShortestEdge','Rot90','ColorJitter','RandomFliplr','RandomFlipud'
+    'ResizeShortestEdge', 'BioMedicalGaussianNoise', 'BioMedicalGaussianBlur',
+    'BioMedicalRandomGamma'
 ]

mmseg/datasets/transforms/transforms.py

@@ -2,17 +2,15 @@
 import copy
 import warnings
 from typing import Dict, Sequence, Tuple, Union
-from PIL import Image
+
 import cv2
 import mmcv
 import numpy as np
 from mmcv.transforms.base import BaseTransform
 from mmcv.transforms.utils import cache_randomness
 from mmengine.utils import is_tuple_of
 from numpy import random
-from torchvision import transforms
-import imgaug.augmenters as iaa
-import imgaug as ia
+from scipy.ndimage import gaussian_filter
 
 from mmseg.datasets.dataset_wrappers import MultiImageMixDataset
 from mmseg.registry import TRANSFORMS
@@ -86,212 +84,6 @@ def __repr__(self):
         return repr_str
 
 
-@TRANSFORMS.register_module()
-class Rot90(BaseTransform):
-    """rotate img and gt with 90,180 or 270 degree.
-
-    Required Keys:
-
-    - img
-    - gt_seg_map
-
-    Modified Keys:
-    - img
-    - seg_fields
-
-    Args:
-        degree_range (tuple): degree range for rotate.
-
-    """
-
-    def __init__(self, degree_range: tuple):
-        self.degree_range = degree_range
-        # self.interpolation = interpolation
-        self.rot90 = iaa.Rot90(degree_range)
-
-    def transform(self, results: dict) -> dict:
-        """rotate img and gt with 90,180 or 270 degree.
-        Args:
-            results (dict): Result dict from loading pipeline.
-
-        Returns:
-            dict: Rotated results.
-        """
-        # Align image to multiple of size divisor.
-        img = results['img']
-
-        gt_segs = []
-        # Align segmentation map to multiple of size divisor.
-        for key in results.get('seg_fields', []):
-            gt_segs.append(results[key])
-        imgs = [img]+gt_segs
-        roted_imgs = self.rot90.augment_images(imgs)
-        results['img'] = roted_imgs[0].astype(np.float32)
-        for i,key in enumerate(results.get('seg_fields', [])):
-            results[key] = roted_imgs[i+1].astype(np.float32)
-
-        return results
-
-    def __repr__(self):
-        repr_str = self.__class__.__name__
-        repr_str += (f'(degree_range={self.degree_range})')
-        return repr_str
-
-
-@TRANSFORMS.register_module()
-class RandomFliplr(BaseTransform):
-    """flip img and gt Horizontally.
-
-    Required Keys:
-
-    - img
-    - gt_seg_map
-
-    Modified Keys:
-    - img
-    - seg_fields
-
-    Args:
-        prob (float): probability for flip Horizontally.
-
-    """
-
-    def __init__(self, prob: float):
-        self.prob = prob
-        self.fliplr = iaa.Fliplr(prob)
-
-    def transform(self, results: dict) -> dict:
-        """flip img and gt Horizontally.
-        Args:
-            results (dict): Result dict from loading pipeline.
-
-        Returns:
-            dict: Flipped results.
-        """
-        img = results['img']
-
-        gt_segs = []
-
-        for key in results.get('seg_fields', []):
-            gt_segs.append(results[key])
-        imgs = [img] + gt_segs
-        flipped_imgs = self.fliplr.augment_images(imgs)
-        results['img'] = flipped_imgs[0].astype(np.float32)
-        for i, key in enumerate(results.get('seg_fields', [])):
-            results[key] = flipped_imgs[i + 1].astype(np.float32)
-
-        return results
-
-    def __repr__(self):
-        repr_str = self.__class__.__name__
-        repr_str += (f'(prob={self.prob})')
-        return repr_str
-
-@TRANSFORMS.register_module()
-class RandomFlipud(BaseTransform):
-    """flip img and gt vertivally.
-
-    Required Keys:
-
-    - img
-    - gt_seg_map
-
-    Modified Keys:
-    - img
-    - seg_fields
-
-    Args:
-        prob (float): probability for flip vertivally.
-
-    """
-
-    def __init__(self, prob: float):
-        self.prob = prob
-        self.flipud = iaa.Flipud(prob)
-
-    def transform(self, results: dict) -> dict:
-        """flip img and gt Horizontally.
-        Args:
-            results (dict): Result dict from loading pipeline.
-
-        Returns:
-            dict: Flipped results.
-        """
-        img = results['img']
-
-        gt_segs = []
-
-        for key in results.get('seg_fields', []):
-            gt_segs.append(results[key])
-        imgs = [img] + gt_segs
-        flipped_imgs = self.flipud.augment_images(imgs)
-        results['img'] = flipped_imgs[0].astype(np.float32)
-        for i, key in enumerate(results.get('seg_fields', [])):
-            results[key] = flipped_imgs[i + 1].astype(np.float32)
-
-        return results
-
-    def __repr__(self):
-        repr_str = self.__class__.__name__
-        repr_str += (f'(prob={self.prob})')
-        return repr_str
-
-
-@TRANSFORMS.register_module()
-class ColorJitter(BaseTransform):
-    """transform img's brightness,contrast,saturation and hue.
-
-    Required Keys:
-    - img
-
-    Modified Keys:
-    - img
-    Args:
-        brightness (float): brightness.
-        contrast (float): contrast.
-        saturation (float): saturation.
-        hue (float): hue.
-
-    """
-
-    def __init__(self, brightness: float=None,contrast: float=None,saturation: float=None,hue: float=None):
-        self.brightness = brightness
-        self.contrast = contrast
-        self.saturation = saturation
-        self.hue = hue
-        color_params = ['brightness','contrast','saturation','hue']
-        self.color_params = {}
-        for param in color_params:
-            if eval(param) is not None:
-                self.color_params.update({param:eval(param)})
-        self.color_jitter = transforms.ColorJitter(**self.color_params)
-
-    def transform(self, results: dict) -> dict:
-        """transform img's brightness,contrast,saturation and hue .
-        Args:
-            results (dict): Result dict from loading pipeline.
-
-        Returns:
-            dict: transformed results.
-        """
-
-        img = results['img']
-        img = Image.fromarray(np.uint8(img))
-        aug_img = self.color_jitter(img)
-        results['img'] = np.array(aug_img,dtype=np.float32)
-
-        return results
-
-    def __repr__(self):
-        repr_str = self.__class__.__name__
-        repr_str += (f'(brightness={self.brightness},'
-                     f'contrast={self.contrast},'
-                     f'saturation={self.saturation},'
-                     f'hue={self.hue})')
-        return repr_str
-
-
-
 @TRANSFORMS.register_module()
 class Rerange(BaseTransform):
     """Rerange the image pixel value.