[Feature] Support ISPRS Vaihingen Dataset. (#1171)

* Add Vaihingen

* upload models&logs of vaihingen

* fix unit test

* fix dataset pipeline

* fix unit test coverage

* fix vaihingen docstring

configs/_base_/datasets/vaihingen.py

@@ -0,0 +1,54 @@
+# dataset settings
+dataset_type = 'ISPRSDataset'
+data_root = 'data/vaihingen'
+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=(512, 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=(512, 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='img_dir/train',
+        ann_dir='ann_dir/train',
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='img_dir/val',
+        ann_dir='ann_dir/val',
+        pipeline=test_pipeline),
+    test=dict(
+        type=dataset_type,
+        data_root=data_root,
+        img_dir='img_dir/val',
+        ann_dir='ann_dir/val',
+        pipeline=test_pipeline))

configs/deeplabv3plus/README.md

@@ -112,6 +112,14 @@ Spatial pyramid pooling module or encode-decoder structure are used in deep neur
 | DeepLabV3+ | R-50-D8  | 512x512   |   80000 | 7.36      | 26.44             | 78.33 |         79.27 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3plus/deeplabv3plus_r50-d8_512x512_80k_potsdam.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r50-d8_512x512_80k_potsdam/deeplabv3plus_r50-d8_512x512_80k_potsdam_20211219_031508-7e7a2b24.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r50-d8_512x512_80k_potsdam/deeplabv3plus_r50-d8_512x512_80k_potsdam_20211219_031508.log.json)     |
 | DeepLabV3+ | R-101-D8 | 512x512   |   80000 | 10.83       | 17.56          | 78.7 |         79.47 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3plus/deeplabv3plus_r101-d8_512x512_80k_potsdam.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_512x512_80k_potsdam/deeplabv3plus_r101-d8_512x512_80k_potsdam_20211219_031508-8b112708.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_512x512_80k_potsdam/deeplabv3plus_r101-d8_512x512_80k_potsdam_20211219_031508.log.json) |
 
+### Vaihingen
+
+| Method     | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) |  mIoU | mIoU(ms+flip) | config                                                                                                                                 | download                                                                                                                                                                                                                                                                                                                                                                                   |
+| ---------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | -------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| DeepLabV3+ | R-18-D8  | 512x512   |   80000 | 1.91      | 72.79             | 72.50 |         74.13 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3plus/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen_20211231_230805-7626a263.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen_20211231_230805.log.json)     |
+| DeepLabV3+ | R-50-D8  | 512x512   |   80000 | 7.36      | 26.91             | 73.97 |         75.05 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3plus/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen_20211231_230816-5040938d.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen_20211231_230816.log.json)     |
+| DeepLabV3+ | R-101-D8 | 512x512   |   80000 | 10.83       | 18.59          | 73.06 |         74.14 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3plus/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen_20211231_230816-8a095afa.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen_20211231_230816.log.json) |
+
 Note:
 
 - `FP16` means Mixed Precision (FP16) is adopted in training.

configs/deeplabv3plus/deeplabv3plus.yml

@@ -9,6 +9,7 @@ Collections:
     - Pascal Context 59
     - LoveDA
     - Potsdam
+    - Vaihingen
   Paper:
     URL: https://arxiv.org/abs/1802.02611
     Title: Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
@@ -736,3 +737,69 @@ Models:
       mIoU(ms+flip): 79.47
   Config: configs/deeplabv3plus/deeplabv3plus_r101-d8_512x512_80k_potsdam.py
   Weights: https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_512x512_80k_potsdam/deeplabv3plus_r101-d8_512x512_80k_potsdam_20211219_031508-8b112708.pth
+- Name: deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen
+  In Collection: deeplabv3plus
+  Metadata:
+    backbone: R-18-D8
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 13.74
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 1.91
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 72.5
+      mIoU(ms+flip): 74.13
+  Config: configs/deeplabv3plus/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen_20211231_230805-7626a263.pth
+- Name: deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen
+  In Collection: deeplabv3plus
+  Metadata:
+    backbone: R-50-D8
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 37.16
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 7.36
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 73.97
+      mIoU(ms+flip): 75.05
+  Config: configs/deeplabv3plus/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen_20211231_230816-5040938d.pth
+- Name: deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen
+  In Collection: deeplabv3plus
+  Metadata:
+    backbone: R-101-D8
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 53.79
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 10.83
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 73.06
+      mIoU(ms+flip): 74.14
+  Config: configs/deeplabv3plus/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen_20211231_230816-8a095afa.pth

configs/deeplabv3plus/deeplabv3plus_r101-d8_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,2 @@
+_base_ = './deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen.py'
+model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))

configs/deeplabv3plus/deeplabv3plus_r18-d8_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,11 @@
+_base_ = './deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen.py'
+model = dict(
+    pretrained='open-mmlab://resnet18_v1c',
+    backbone=dict(depth=18),
+    decode_head=dict(
+        c1_in_channels=64,
+        c1_channels=12,
+        in_channels=512,
+        channels=128,
+    ),
+    auxiliary_head=dict(in_channels=256, channels=64))

configs/deeplabv3plus/deeplabv3plus_r50-d8_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,7 @@
+_base_ = [
+    '../_base_/models/deeplabv3plus_r50-d8.py',
+    '../_base_/datasets/vaihingen.py', '../_base_/default_runtime.py',
+    '../_base_/schedules/schedule_80k.py'
+]
+model = dict(
+    decode_head=dict(num_classes=6), auxiliary_head=dict(num_classes=6))

configs/hrnet/README.md

@@ -100,3 +100,11 @@ High-resolution representations are essential for position-sensitive vision prob
 | FCN | HRNetV2p-W18-Small  | 512x512   |   80000 | 1.58      | 36.00             | 77.64 |         78.8 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/hrnet/fcn_hr18s_512x512_80k_potsdam.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18s_512x512_80k_potsdam/fcn_hr18s_512x512_80k_potsdam_20211218_205517-ba32af63.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18s_512x512_80k_potsdam/fcn_hr18s_512x512_80k_potsdam_20211218_205517.log.json)     |
 | FCN | HRNetV2p-W18  | 512x512   |   80000 | 2.76      | 19.25             | 78.26 |         79.24 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/hrnet/fcn_hr18_512x512_80k_potsdam.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18_512x512_80k_potsdam/fcn_hr18_512x512_80k_potsdam_20211218_205517-5d0387ad.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18_512x512_80k_potsdam/fcn_hr18_512x512_80k_potsdam_20211218_205517.log.json)     |
 | FCN | HRNetV2p-W48 | 512x512   |   80000 | 6.20       | 16.42          | 78.39 |         79.34 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/hrnet/fcn_hr48_512x512_80k_potsdam.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr48_512x512_80k_potsdam/fcn_hr48_512x512_80k_potsdam_20211219_020601-97434c78.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr48_512x512_80k_potsdam/fcn_hr48_512x512_80k_potsdam_20211219_020601.log.json) |
+
+### Vaihingen
+
+| Method     | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) |  mIoU | mIoU(ms+flip) | config                                                                                                                                 | download                                                                                                                                                                                                                                                                                                                                                                                   |
+| ---------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | -------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| FCN | HRNetV2p-W18-Small  | 512x512   |   80000 | 1.58      | 38.11             | 71.81 |         73.1 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/hrnet/fcn_hr18s_4x4_512x512_80k_vaihingen.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18s_4x4_512x512_80k_vaihingen/fcn_hr18s_4x4_512x512_80k_vaihingen_20211231_230909-b23aae02.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18s_4x4_512x512_80k_vaihingen/fcn_hr18s_4x4_512x512_80k_vaihingen_20211231_230909.log.json)     |
+| FCN | HRNetV2p-W18  | 512x512   |   80000 | 2.76      | 19.55             | 72.57 |         74.09 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/hrnet/fcn_hr18_4x4_512x512_80k_vaihingen.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18_4x4_512x512_80k_vaihingen/fcn_hr18_4x4_512x512_80k_vaihingen_20211231_231216-2ec3ae8a.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18_4x4_512x512_80k_vaihingen/fcn_hr18_4x4_512x512_80k_vaihingen_20211231_231216.log.json)     |
+| FCN | HRNetV2p-W48 | 512x512   |   80000 | 6.20       | 17.25          | 72.50 |         73.52 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/hrnet/fcn_hr48_4x4_512x512_80k_vaihingen.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr48_4x4_512x512_80k_vaihingen/fcn_hr48_4x4_512x512_80k_vaihingen_20211231_231244-7133cb22.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr48_4x4_512x512_80k_vaihingen/fcn_hr48_4x4_512x512_80k_vaihingen_20211231_231244.log.json) |

configs/hrnet/fcn_hr18_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,5 @@
+_base_ = [
+    '../_base_/models/fcn_hr18.py', '../_base_/datasets/vaihingen.py',
+    '../_base_/default_runtime.py', '../_base_/schedules/schedule_80k.py'
+]
+model = dict(decode_head=dict(num_classes=6))

configs/hrnet/fcn_hr18s_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,9 @@
+_base_ = './fcn_hr18_4x4_512x512_80k_vaihingen.py'
+model = dict(
+    pretrained='open-mmlab://msra/hrnetv2_w18_small',
+    backbone=dict(
+        extra=dict(
+            stage1=dict(num_blocks=(2, )),
+            stage2=dict(num_blocks=(2, 2)),
+            stage3=dict(num_modules=3, num_blocks=(2, 2, 2)),
+            stage4=dict(num_modules=2, num_blocks=(2, 2, 2, 2)))))

configs/hrnet/fcn_hr48_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,10 @@
+_base_ = './fcn_hr18_4x4_512x512_80k_vaihingen.py'
+model = dict(
+    pretrained='open-mmlab://msra/hrnetv2_w48',
+    backbone=dict(
+        extra=dict(
+            stage2=dict(num_channels=(48, 96)),
+            stage3=dict(num_channels=(48, 96, 192)),
+            stage4=dict(num_channels=(48, 96, 192, 384)))),
+    decode_head=dict(
+        in_channels=[48, 96, 192, 384], channels=sum([48, 96, 192, 384])))

configs/hrnet/hrnet.yml

@@ -9,6 +9,7 @@ Collections:
     - Pascal Context 59
     - LoveDA
     - Potsdam
+    - Vaihingen
   Paper:
     URL: https://arxiv.org/abs/1908.07919
     Title: Deep High-Resolution Representation Learning for Human Pose Estimation
@@ -581,3 +582,69 @@ Models:
       mIoU(ms+flip): 79.34
   Config: configs/hrnet/fcn_hr48_512x512_80k_potsdam.py
   Weights: https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr48_512x512_80k_potsdam/fcn_hr48_512x512_80k_potsdam_20211219_020601-97434c78.pth
+- Name: fcn_hr18s_4x4_512x512_80k_vaihingen
+  In Collection: hrnet
+  Metadata:
+    backbone: HRNetV2p-W18-Small
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 26.24
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 1.58
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 71.81
+      mIoU(ms+flip): 73.1
+  Config: configs/hrnet/fcn_hr18s_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18s_4x4_512x512_80k_vaihingen/fcn_hr18s_4x4_512x512_80k_vaihingen_20211231_230909-b23aae02.pth
+- Name: fcn_hr18_4x4_512x512_80k_vaihingen
+  In Collection: hrnet
+  Metadata:
+    backbone: HRNetV2p-W18
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 51.15
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 2.76
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 72.57
+      mIoU(ms+flip): 74.09
+  Config: configs/hrnet/fcn_hr18_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr18_4x4_512x512_80k_vaihingen/fcn_hr18_4x4_512x512_80k_vaihingen_20211231_231216-2ec3ae8a.pth
+- Name: fcn_hr48_4x4_512x512_80k_vaihingen
+  In Collection: hrnet
+  Metadata:
+    backbone: HRNetV2p-W48
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 57.97
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 6.2
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 72.5
+      mIoU(ms+flip): 73.52
+  Config: configs/hrnet/fcn_hr48_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/hrnet/fcn_hr48_4x4_512x512_80k_vaihingen/fcn_hr48_4x4_512x512_80k_vaihingen_20211231_231244-7133cb22.pth

configs/pspnet/README.md

@@ -141,6 +141,14 @@ We support evaluation results on these two datasets using models above trained o
 | PSPNet | R-50-D8  | 512x512   |   80000 | 6.14      | 30.21             | 78.12 |         78.98 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/pspnet/pspnet_r50-d8_4x4_512x512_80k_potsdam.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r50-d8_4x4_512x512_80k_potsdam/pspnet_r50-d8_4x4_512x512_80k_potsdam_20211219_043541-2dd5fe67.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r50-d8_4x4_512x512_80k_potsdam/pspnet_r50-d8_4x4_512x512_80k_potsdam_20211219_043541.log.json)     |
 | PSPNet | R-101-D8 | 512x512   |   80000 | 9.61       | 19.40          | 78.62 |         79.47 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/pspnet/pspnet_r101-d8_4x4_512x512_80k_potsdam.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_4x4_512x512_80k_potsdam/pspnet_r101-d8_4x4_512x512_80k_potsdam_20211220_125612-aed036c4.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_4x4_512x512_80k_potsdam/pspnet_r101-d8_4x4_512x512_80k_potsdam_20211220_125612.log.json) |
 
+### Vaihingen
+
+| Method     | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) |  mIoU | mIoU(ms+flip) | config                                                                                                                                 | download                                                                                                                                                                                                                                                                                                                                                                                   |
+| ---------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | -------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| PSPNet | R-18-D8  | 512x512   |   80000 | 1.45      | 85.06             | 71.46 |         73.36 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/pspnet/pspnet_r18-d8_4x4_512x512_80k_vaihingen.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r18-d8_4x4_512x512_80k_vaihingen/pspnet_r18-d8_4x4_512x512_80k_vaihingen_20211228_160355-52a8a6f6.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r18-d8_4x4_512x512_80k_vaihingen/pspnet_r18-d8_4x4_512x512_80k_vaihingen_20211228_160355.log.json)     |
+| PSPNet | R-50-D8  | 512x512   |   80000 | 6.14      | 30.29             | 72.36 |         73.75 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/pspnet/pspnet_r50-d8_4x4_512x512_80k_vaihingen.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r50-d8_4x4_512x512_80k_vaihingen/pspnet_r50-d8_4x4_512x512_80k_vaihingen_20211228_160355-382f8f5b.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r50-d8_4x4_512x512_80k_vaihingen/pspnet_r50-d8_4x4_512x512_80k_vaihingen_20211228_160355.log.json)     |
+| PSPNet | R-101-D8 | 512x512   |   80000 | 9.61       | 19.97          | 72.61 |         74.18 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/pspnet/pspnet_r101-d8_4x4_512x512_80k_vaihingen.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_4x4_512x512_80k_vaihingen/pspnet_r101-d8_4x4_512x512_80k_vaihingen_20211231_230806-8eba0a09.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_4x4_512x512_80k_vaihingen/pspnet_r101-d8_4x4_512x512_80k_vaihingen_20211231_230806.log.json) |
+
 Note:
 
 - `FP16` means Mixed Precision (FP16) is adopted in training.

configs/pspnet/pspnet.yml

@@ -12,6 +12,7 @@ Collections:
     - COCO-Stuff 164k
     - LoveDA
     - Potsdam
+    - Vaihingen
   Paper:
     URL: https://arxiv.org/abs/1612.01105
     Title: Pyramid Scene Parsing Network
@@ -875,3 +876,69 @@ Models:
       mIoU(ms+flip): 79.47
   Config: configs/pspnet/pspnet_r101-d8_4x4_512x512_80k_potsdam.py
   Weights: https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_4x4_512x512_80k_potsdam/pspnet_r101-d8_4x4_512x512_80k_potsdam_20211220_125612-aed036c4.pth
+- Name: pspnet_r18-d8_4x4_512x512_80k_vaihingen
+  In Collection: pspnet
+  Metadata:
+    backbone: R-18-D8
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 11.76
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 1.45
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 71.46
+      mIoU(ms+flip): 73.36
+  Config: configs/pspnet/pspnet_r18-d8_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r18-d8_4x4_512x512_80k_vaihingen/pspnet_r18-d8_4x4_512x512_80k_vaihingen_20211228_160355-52a8a6f6.pth
+- Name: pspnet_r50-d8_4x4_512x512_80k_vaihingen
+  In Collection: pspnet
+  Metadata:
+    backbone: R-50-D8
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 33.01
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 6.14
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 72.36
+      mIoU(ms+flip): 73.75
+  Config: configs/pspnet/pspnet_r50-d8_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r50-d8_4x4_512x512_80k_vaihingen/pspnet_r50-d8_4x4_512x512_80k_vaihingen_20211228_160355-382f8f5b.pth
+- Name: pspnet_r101-d8_4x4_512x512_80k_vaihingen
+  In Collection: pspnet
+  Metadata:
+    backbone: R-101-D8
+    crop size: (512,512)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 50.08
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,512)
+    Training Memory (GB): 9.61
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: Vaihingen
+    Metrics:
+      mIoU: 72.61
+      mIoU(ms+flip): 74.18
+  Config: configs/pspnet/pspnet_r101-d8_4x4_512x512_80k_vaihingen.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_4x4_512x512_80k_vaihingen/pspnet_r101-d8_4x4_512x512_80k_vaihingen_20211231_230806-8eba0a09.pth

configs/pspnet/pspnet_r101-d8_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,2 @@
+_base_ = './pspnet_r50-d8_4x4_512x512_80k_vaihingen.py'
+model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))

configs/pspnet/pspnet_r18-d8_4x4_512x512_80k_vaihingen.py

@@ -0,0 +1,9 @@
+_base_ = './pspnet_r50-d8_4x4_512x512_80k_vaihingen.py'
+model = dict(
+    pretrained='open-mmlab://resnet18_v1c',
+    backbone=dict(depth=18),
+    decode_head=dict(
+        in_channels=512,
+        channels=128,
+    ),
+    auxiliary_head=dict(in_channels=256, channels=64))

configs/pspnet/pspnet_r50-d8_4x4_512x512_80k_vaihingen.py

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

docs/en/dataset_prepare.md

@@ -309,3 +309,19 @@ python tools/convert_datasets/potsdam.py /path/to/potsdam
 ```
 
 In our default setting, it will generate 3456 images for training and 2016 images for validation.
+
+### ISPRS Vaihingen
+
+The [Vaihingen](https://www2.isprs.org/commissions/comm2/wg4/benchmark/2d-sem-label-vaihingen/)
+dataset is for urban semantic segmentation used in the 2D Semantic Labeling Contest - Vaihingen.
+
+The dataset can be requested at the challenge [homepage](https://www2.isprs.org/commissions/comm2/wg4/benchmark/data-request-form/).
+The 'ISPRS_semantic_labeling_Vaihingen.zip' and 'ISPRS_semantic_labeling_Vaihingen_ground_truth_eroded_COMPLETE.zip' are required.
+
+For Vaihingen dataset, please run the following command to download and re-organize the dataset.
+
+```shell
+python tools/convert_datasets/vaihingen.py /path/to/vaihingen
+```
+
+In our default setting (`clip_size` =512, `stride_size`=256), it will generate 344 images for training and 398 images for validation.

docs/zh_cn/dataset_prepare.md

@@ -250,3 +250,19 @@ python tools/convert_datasets/potsdam.py /path/to/potsdam
 ```
 
 使用我们默认的配置， 将生成 3456 张图片的训练集和 2016 张图片的验证集。
+
+### ISPRS Vaihingen
+
+[Vaihingen](https://www2.isprs.org/commissions/comm2/wg4/benchmark/2d-sem-label-vaihingen/)
+数据集是一个有着2D 语义分割内容标注的城市遥感数据集。
+
+数据集可以从挑战 [主页](https://www2.isprs.org/commissions/comm2/wg4/benchmark/data-request-form/).
+需要其中的 'ISPRS_semantic_labeling_Vaihingen.zip' 和 'ISPRS_semantic_labeling_Vaihingen_ground_truth_eroded_COMPLETE.zip'。
+
+对于 Vaihingen 数据集，请运行以下命令下载并重新组织数据集
+
+```shell
+python tools/convert_datasets/vaihingen.py /path/to/vaihingen
+```
+
+使用我们默认的配置 (`clip_size` =512, `stride_size`=256)， 将生成 344 张图片的训练集和 398 张图片的验证集。

mmseg/core/evaluation/class_names.py

@@ -52,22 +52,6 @@ def voc_classes():
     ]
 
 
-def loveda_classes():
-    """LoveDA class names for external use."""
-    return [
-        'background', 'building', 'road', 'water', 'barren', 'forest',
-        'agricultural'
-    ]
-
-
-def potsdam_classes():
-    """Potsdam class names for external use."""
-    return [
-        'impervious_surface', 'building', 'low_vegetation', 'tree', 'car',
-        'clutter'
-    ]
-
-
 def cocostuff_classes():
     """CocoStuff class names for external use."""
     return [
@@ -103,6 +87,30 @@ def cocostuff_classes():
     ]
 
 
+def loveda_classes():
+    """LoveDA class names for external use."""
+    return [
+        'background', 'building', 'road', 'water', 'barren', 'forest',
+        'agricultural'
+    ]
+
+
+def potsdam_classes():
+    """Potsdam class names for external use."""
+    return [
+        'impervious_surface', 'building', 'low_vegetation', 'tree', 'car',
+        'clutter'
+    ]
+
+
+def vaihingen_classes():
+    """Vaihingen class names for external use."""
+    return [
+        'impervious_surface', 'building', 'low_vegetation', 'tree', 'car',
+        'clutter'
+    ]
+
+
 def cityscapes_palette():
     """Cityscapes palette for external use."""
     return [[128, 64, 128], [244, 35, 232], [70, 70, 70], [102, 102, 156],
@@ -163,18 +171,6 @@ def voc_palette():
             [128, 64, 0], [0, 192, 0], [128, 192, 0], [0, 64, 128]]
 
 
-def loveda_palette():
-    """LoveDA palette for external use."""
-    return [[255, 255, 255], [255, 0, 0], [255, 255, 0], [0, 0, 255],
-            [159, 129, 183], [0, 255, 0], [255, 195, 128]]
-
-
-def potsdam_palette():
-    """Potsdam palette for external use."""
-    return [[255, 255, 255], [0, 0, 255], [0, 255, 255], [0, 255, 0],
-            [255, 255, 0], [255, 0, 0]]
-
-
 def cocostuff_palette():
     """CocoStuff palette for external use."""
     return [[0, 192, 64], [0, 192, 64], [0, 64, 96], [128, 192, 192],
@@ -222,12 +218,31 @@ def cocostuff_palette():
             [64, 160, 64], [64, 64, 0]]
 
 
+def loveda_palette():
+    """LoveDA palette for external use."""
+    return [[255, 255, 255], [255, 0, 0], [255, 255, 0], [0, 0, 255],
+            [159, 129, 183], [0, 255, 0], [255, 195, 128]]
+
+
+def potsdam_palette():
+    """Potsdam palette for external use."""
+    return [[255, 255, 255], [0, 0, 255], [0, 255, 255], [0, 255, 0],
+            [255, 255, 0], [255, 0, 0]]
+
+
+def vaihingen_palette():
+    """Vaihingen palette for external use."""
+    return [[255, 255, 255], [0, 0, 255], [0, 0, 255], [0, 255, 0],
+            [255, 255, 0], [255, 0, 0]]
+
+
 dataset_aliases = {
     'cityscapes': ['cityscapes'],
     'ade': ['ade', 'ade20k'],
     'voc': ['voc', 'pascal_voc', 'voc12', 'voc12aug'],
     'loveda': ['loveda'],
     'potsdam': ['potsdam'],
+    'vaihingen': ['vaihingen'],
     'cocostuff': [
         'cocostuff', 'cocostuff10k', 'cocostuff164k', 'coco-stuff',
         'coco-stuff10k', 'coco-stuff164k', 'coco_stuff', 'coco_stuff10k',

mmseg/datasets/__init__.py

@@ -10,6 +10,7 @@ from .dataset_wrappers import (ConcatDataset, MultiImageMixDataset,
                                RepeatDataset)
 from .drive import DRIVEDataset
 from .hrf import HRFDataset
+from .isprs import ISPRSDataset
 from .loveda import LoveDADataset
 from .night_driving import NightDrivingDataset
 from .pascal_context import PascalContextDataset, PascalContextDataset59
@@ -24,5 +25,5 @@ __all__ = [
     'PascalContextDataset59', 'ChaseDB1Dataset', 'DRIVEDataset', 'HRFDataset',
     'STAREDataset', 'DarkZurichDataset', 'NightDrivingDataset',
     'COCOStuffDataset', 'LoveDADataset', 'MultiImageMixDataset',
-    'PotsdamDataset'
+    'ISPRSDataset', 'PotsdamDataset'
 ]

mmseg/datasets/isprs.py

@@ -0,0 +1,25 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+from .builder import DATASETS
+from .custom import CustomDataset
+
+
+@DATASETS.register_module()
+class ISPRSDataset(CustomDataset):
+    """ISPRS dataset.
+
+    In segmentation map annotation for LoveDA, 0 is the ignore index.
+    ``reduce_zero_label`` should be set to True. The ``img_suffix`` and
+    ``seg_map_suffix`` are both fixed to '.png'.
+    """
+    CLASSES = ('impervious_surface', 'building', 'low_vegetation', 'tree',
+               'car', 'clutter')
+
+    PALETTE = [[255, 255, 255], [0, 0, 255], [0, 0, 255], [0, 255, 0],
+               [255, 255, 0], [255, 0, 0]]
+
+    def __init__(self, **kwargs):
+        super(ISPRSDataset, self).__init__(
+            img_suffix='.png',
+            seg_map_suffix='.png',
+            reduce_zero_label=True,
+            **kwargs)

tests/test_data/test_dataset.py

@@ -14,7 +14,7 @@ from PIL import Image
 from mmseg.core.evaluation import get_classes, get_palette
 from mmseg.datasets import (DATASETS, ADE20KDataset, CityscapesDataset,
                             COCOStuffDataset, ConcatDataset, CustomDataset,
-                            LoveDADataset, MultiImageMixDataset,
+                            ISPRSDataset, LoveDADataset, MultiImageMixDataset,
                             PascalVOCDataset, PotsdamDataset, RepeatDataset,
                             build_dataset)
 
@@ -27,6 +27,7 @@ def test_classes():
         ADE20KDataset.CLASSES) == get_classes('ade') == get_classes('ade20k')
     assert list(LoveDADataset.CLASSES) == get_classes('loveda')
     assert list(PotsdamDataset.CLASSES) == get_classes('potsdam')
+    assert list(ISPRSDataset.CLASSES) == get_classes('vaihingen')
     assert list(COCOStuffDataset.CLASSES) == get_classes('cocostuff')
 
     with pytest.raises(ValueError):
@@ -719,6 +720,16 @@ def test_potsdam():
     assert len(test_dataset) == 1
 
 
+def test_vaihingen():
+    test_dataset = ISPRSDataset(
+        pipeline=[],
+        img_dir=osp.join(
+            osp.dirname(__file__), '../data/pseudo_vaihingen_dataset/img_dir'),
+        ann_dir=osp.join(
+            osp.dirname(__file__), '../data/pseudo_vaihingen_dataset/ann_dir'))
+    assert len(test_dataset) == 1
+
+
 @patch('mmseg.datasets.CustomDataset.load_annotations', MagicMock)
 @patch('mmseg.datasets.CustomDataset.__getitem__',
        MagicMock(side_effect=lambda idx: idx))

tools/convert_datasets/vaihingen.py

@@ -0,0 +1,155 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import argparse
+import glob
+import math
+import os
+import os.path as osp
+import tempfile
+import zipfile
+
+import mmcv
+import numpy as np
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(
+        description='Convert vaihingen dataset to mmsegmentation format')
+    parser.add_argument('dataset_path', help='vaihingen folder path')
+    parser.add_argument('--tmp_dir', help='path of the temporary directory')
+    parser.add_argument('-o', '--out_dir', help='output path')
+    parser.add_argument(
+        '--clip_size',
+        type=int,
+        help='clipped size of image after preparation',
+        default=512)
+    parser.add_argument(
+        '--stride_size',
+        type=int,
+        help='stride of clipping original images',
+        default=256)
+    args = parser.parse_args()
+    return args
+
+
+def clip_big_image(image_path, clip_save_dir, to_label=False):
+    # Original image of Vaihingen dataset is very large, thus pre-processing
+    # of them is adopted. Given fixed clip size and stride size to generate
+    # clipped image, the intersection　of width and height is determined.
+    # For example, given one 5120 x 5120 original image, the clip size is
+    # 512 and stride size is 256, thus it would generate 20x20 = 400 images
+    # whose size are all 512x512.
+    image = mmcv.imread(image_path)
+
+    h, w, c = image.shape
+    cs = args.clip_size
+    ss = args.stride_size
+
+    num_rows = math.ceil((h - cs) / ss) if math.ceil(
+        (h - cs) / ss) * ss + cs >= h else math.ceil((h - cs) / ss) + 1
+    num_cols = math.ceil((w - cs) / ss) if math.ceil(
+        (w - cs) / ss) * ss + cs >= w else math.ceil((w - cs) / ss) + 1
+
+    x, y = np.meshgrid(np.arange(num_cols + 1), np.arange(num_rows + 1))
+    xmin = x * cs
+    ymin = y * cs
+
+    xmin = xmin.ravel()
+    ymin = ymin.ravel()
+    xmin_offset = np.where(xmin + cs > w, w - xmin - cs, np.zeros_like(xmin))
+    ymin_offset = np.where(ymin + cs > h, h - ymin - cs, np.zeros_like(ymin))
+    boxes = np.stack([
+        xmin + xmin_offset, ymin + ymin_offset,
+        np.minimum(xmin + cs, w),
+        np.minimum(ymin + cs, h)
+    ],
+                     axis=1)
+
+    if to_label:
+        color_map = np.array([[0, 0, 0], [255, 255, 255], [255, 0, 0],
+                              [255, 255, 0], [0, 255, 0], [0, 255, 255],
+                              [0, 0, 255]])
+        flatten_v = np.matmul(
+            image.reshape(-1, c),
+            np.array([2, 3, 4]).reshape(3, 1))
+        out = np.zeros_like(flatten_v)
+        for idx, class_color in enumerate(color_map):
+            value_idx = np.matmul(class_color,
+                                  np.array([2, 3, 4]).reshape(3, 1))
+            out[flatten_v == value_idx] = idx
+        image = out.reshape(h, w)
+
+    for box in boxes:
+        start_x, start_y, end_x, end_y = box
+        clipped_image = image[start_y:end_y,
+                              start_x:end_x] if to_label else image[
+                                  start_y:end_y, start_x:end_x, :]
+        area_idx = osp.basename(image_path).split('_')[3].strip('.tif')
+        mmcv.imwrite(
+            clipped_image.astype(np.uint8),
+            osp.join(clip_save_dir,
+                     f'{area_idx}_{start_x}_{start_y}_{end_x}_{end_y}.png'))
+
+
+def main():
+    splits = {
+        'train': [
+            'area1', 'area11', 'area13', 'area15', 'area17', 'area21',
+            'area23', 'area26', 'area28', 'area3', 'area30', 'area32',
+            'area34', 'area37', 'area5', 'area7'
+        ],
+        'val': [
+            'area6', 'area24', 'area35', 'area16', 'area14', 'area22',
+            'area10', 'area4', 'area2', 'area20', 'area8', 'area31', 'area33',
+            'area27', 'area38', 'area12', 'area29'
+        ],
+    }
+
+    dataset_path = args.dataset_path
+    if args.out_dir is None:
+        out_dir = osp.join('data', 'vaihingen')
+    else:
+        out_dir = args.out_dir
+
+    print('Making directories...')
+    mmcv.mkdir_or_exist(osp.join(out_dir, 'img_dir', 'train'))
+    mmcv.mkdir_or_exist(osp.join(out_dir, 'img_dir', 'val'))
+    mmcv.mkdir_or_exist(osp.join(out_dir, 'ann_dir', 'train'))
+    mmcv.mkdir_or_exist(osp.join(out_dir, 'ann_dir', 'val'))
+
+    zipp_list = glob.glob(os.path.join(dataset_path, '*.zip'))
+    print('Find the data', zipp_list)
+
+    with tempfile.TemporaryDirectory(dir=args.tmp_dir) as tmp_dir:
+        for zipp in zipp_list:
+            zip_file = zipfile.ZipFile(zipp)
+            zip_file.extractall(tmp_dir)
+            src_path_list = glob.glob(os.path.join(tmp_dir, '*.tif'))
+            if 'ISPRS_semantic_labeling_Vaihingen' in zipp:
+                src_path_list = glob.glob(
+                    os.path.join(os.path.join(tmp_dir, 'top'), '*.tif'))
+            if 'ISPRS_semantic_labeling_Vaihingen_ground_truth_eroded_COMPLETE' in zipp:  # noqa
+                src_path_list = glob.glob(os.path.join(tmp_dir, '*.tif'))
+                # delete unused area9 ground truth
+                for area_ann in src_path_list:
+                    if 'area9' in area_ann:
+                        src_path_list.remove(area_ann)
+            prog_bar = mmcv.ProgressBar(len(src_path_list))
+            for i, src_path in enumerate(src_path_list):
+                area_idx = osp.basename(src_path).split('_')[3].strip('.tif')
+                data_type = 'train' if area_idx in splits['train'] else 'val'
+                if 'noBoundary' in src_path:
+                    dst_dir = osp.join(out_dir, 'ann_dir', data_type)
+                    clip_big_image(src_path, dst_dir, to_label=True)
+                else:
+                    dst_dir = osp.join(out_dir, 'img_dir', data_type)
+                    clip_big_image(src_path, dst_dir, to_label=False)
+                prog_bar.update()
+
+        print('Removing the temporary files...')
+
+    print('Done!')
+
+
+if __name__ == '__main__':
+    args = parse_args()
+    main()