Skip to content

Latest commit

 

History

History
185 lines (160 loc) · 4.68 KB

DATA.md

File metadata and controls

185 lines (160 loc) · 4.68 KB
Raw

Prepare Datasets

data_dir
|—— cityscapes
|   |—— leftImg8bit
|   |   |—— train
|   |   |—— val
|   |—— gtFine
|   |   |—— train
|   |   |—— val
|—— NighttimeDrivingTest
|   |—— leftImg8bit
|   |   |—— test
|   |—— gtCoarse_daytime_trainvaltest
|   |   |—— test
|—— dark_zurich
|   |—— rgb_anon
|   |   |—— val
|   |—— gt
|   |   |—— val
|—— mapillary
|   |—— training
|   |   |—— images
|   |   |—— v1.2
|   |—— validation
|   |   |—— images
|   |   |—— v1.2
|—— camvid
|   |—— images
|   |   |—— train
|   |   |—— test
|   |—— annotations
|   |   |—— train
|   |   |—— test
|—— ade
|   |—— ADEChallengeData2016
|   |   |—— images
|   |   |   |—— training
|   |   |   |—— validation
|   |   |—— annotations
|   |   |   |—— training
|   |   |   |—— validation
|—— coco_stuff164k
|   |—— images
|   |   |—— train2017
|   |   |—— val2017
|   |—— annotations
|   |   |—— train2017
|   |   |—— val2017
|—— VOCdevkit
|   |—— VOC2010
|   |   |—— JPEGImages
|   |   |—— SegmentationClassContext
|   |—— VOC2012
|   |   |—— JPEGImages
|   |   |—— SegmentationClass
|   |   |—— SegmentationClassAug
|   |   |—— SegmentationClassTrainAug
|   |—— VOCaug
|—— land
|   |—— train
|—— road
|   |—— train
|—— building
|   |—— train
|—— lits
|   |—— train
|—— kits
|   |—— train
|—— qubiq
|   |—— brain-growth
|   |—— brain-tumor
|   |—— kidney
|   |—— prostate

Many datasets rely on scripts from MMSegmentation. Please refer to here for more details.

Cityscapes

  • Step 1: Download the dataset from here

  • Step 2: Run the following from MMSegmentation

    python tools/dataset_converters/cityscapes.py data/cityscapes

Nighttime Driving

  • Download the test set from here.

Dark Zurich

  • Download the validation set from here.

Mapillary Vistas

  • Step 1: Download the dataset from here
  • Step 2: Run the following 
    python datas/prepare_mapillary_vistas.py path/to/data_dir

CamVid

  • Step 1: Merge the training and validation sets as the folder structure shows.

  • Step 2: Run the following

    python datas/prepare_camvid.py path/to/data_dir train
python datas/prepare_camvid.py path/to/data_dir test

ADE20K

  • Download the dataset from here

COCO-Stuff

  • Run the following from MMSegmentation 
    wget http://images.cocodataset.org/zips/train2017.zip
wget http://images.cocodataset.org/zips/val2017.zip
wget http://calvin.inf.ed.ac.uk/wp-content/uploads/data/cocostuffdataset/stuffthingmaps_trainval2017.zip

unzip train2017.zip -d images/
unzip val2017.zip -d images/
unzip stuffthingmaps_trainval2017.zip -d annotations/

python tools/dataset_converters/coco_stuff164k.py path/to/coco_stuff164k

PASCAL VOC

  • Step 1: Download PASCAL VOC 2012 from here and extra data from here.

  • Step 2: Run the following from MMSegmentation

    python tools/dataset_converters/voc_aug.py \
  /path/to/data_dir/VOCdevkit \
  /path/to/data_dir/VOCdevkit/VOCaug

  • Step 3: Run the following

    python datas/process_pascal_voc.py path/to/data_dir

PASCAL Context

  • Step 1: Install detail following here

  • Step 2: Run the following from MMSegmentation

    python tools/dataset_converters/pascal_context.py data/VOCdevkit data/VOCdevkit/VOC2010/trainval_merged.json

DeepGlobe Land

  • Run the following 
    datas/prepare_deepglobe_land.py path/to/data_dir

DeepGlobe Road

  • Run the following 
    datas/prepare_deepglobe_road.py path/to/data_dir

DeepGlobe Building

  • Run the following 
    datas/prepare_deepglobe_building.py path/to/data_dir

LiTS

  • Run the following 
    datas/prepare_lits_kits.py path/to/data_dir lits

KiTS

  • Run the following 
    datas/prepare_lits_kits.py path/to/data_dir kits

QUBIQ

  • Run the following 
    datas/prepare_qubiq.py path/to/data_dir brain-growth 0