Merge branch 'zhengmiao/randomcrop' into 'refactor_dev'

[Enhancement] Revise RandomCrop

See merge request openmmlab-enterprise/openmmlab-ce/mmsegmentation!20

configs/_base_/datasets/ade20k.py

@@ -8,20 +8,7 @@ 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(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/ade20k_640x640.py

@@ -8,20 +8,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations', reduce_zero_label=True),
     dict(type='Resize', img_scale=(2560, 640), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/chase_db1.py

@@ -9,20 +9,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/cityscapes.py

@@ -8,20 +8,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(2048, 1024), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/cityscapes_1024x1024.py

@@ -6,20 +6,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(2048, 1024), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/cityscapes_768x768.py

@@ -6,20 +6,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(2049, 1025), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/cityscapes_769x769.py

@@ -6,20 +6,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(2049, 1025), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/cityscapes_832x832.py

@@ -6,20 +6,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(2048, 1024), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/coco-stuff10k.py

@@ -8,20 +8,7 @@ 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(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/coco-stuff164k.py

@@ -8,20 +8,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(2048, 512), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/drive.py

@@ -9,20 +9,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/hrf.py

@@ -9,20 +9,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/isaid.py

@@ -16,20 +16,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(896, 896), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/loveda.py

@@ -8,20 +8,7 @@ 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(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/pascal_context.py

@@ -11,20 +11,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/pascal_context_59.py

@@ -11,20 +11,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations', reduce_zero_label=True),
     dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/pascal_voc12.py

@@ -8,20 +8,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=(2048, 512), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/potsdam.py

@@ -8,20 +8,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations', reduce_zero_label=True),
     dict(type='Resize', img_scale=(512, 512), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/stare.py

@@ -9,20 +9,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations'),
     dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

configs/_base_/datasets/vaihingen.py

@@ -8,20 +8,7 @@ train_pipeline = [
     dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations', reduce_zero_label=True),
     dict(type='Resize', img_scale=(512, 512), ratio_range=(0.5, 2.0)),
-    dict(
+    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
-        type='TransformBroadcaster',
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True,
-        transforms=[
-            dict(
-                type='mmseg.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),

mmseg/datasets/pipelines/transforms.py

@@ -237,7 +237,7 @@ class RandomCrop(BaseTransform):
         if self.cat_max_ratio < 1.:
             # Repeat 10 times
             for _ in range(10):
-                seg_temp = self.crop(results['gt_semantic_seg'], crop_bbox)
+                seg_temp = self.crop(results['gt_seg_map'], crop_bbox)
                 labels, cnt = np.unique(seg_temp, return_counts=True)
                 cnt = cnt[labels != self.ignore_index]
                 if len(cnt) > 1 and np.max(cnt) / np.sum(
@@ -279,6 +279,10 @@ class RandomCrop(BaseTransform):
 
         # crop the image
         img = self.crop(img, crop_bbox)
+
+        # crop semantic seg
+        for key in results.get('seg_fields', []):
+            results[key] = self.crop(results[key], crop_bbox)
         img_shape = img.shape
         results['img'] = img
         results['img_shape'] = img_shape

tests/test_datasets/test_pipelines/test_transforms.py

@@ -4,7 +4,6 @@ import os.path as osp
 import mmcv
 import numpy as np
 import pytest
-from mmcv.transforms.wrappers import TransformBroadcaster
 from PIL import Image
 
 from mmseg.datasets.pipelines import PhotoMetricDistortion, RandomCrop
@@ -28,14 +27,8 @@ def test_random_crop():
     results['scale_factor'] = 1.0
 
     h, w, _ = img.shape
-    pipeline = TransformBroadcaster(
+    pipeline = RandomCrop(crop_size=(h - 20, w - 20))
-        transforms=[RandomCrop(crop_size=(h - 20, w - 20))],
+
-        mapping={
-            'img': ['img', 'gt_semantic_seg'],
-            'img_shape': [..., 'img_shape']
-        },
-        auto_remap=True,
-        share_random_params=True)
     results = pipeline(results)
     assert results['img'].shape[:2] == (h - 20, w - 20)
     assert results['img_shape'][:2] == (h - 20, w - 20)