Skip to content

Latest commit

 

History

History
81 lines (69 loc) · 3.87 KB

README.md

File metadata and controls

81 lines (69 loc) · 3.87 KB

Download and preprocess SCARED dataset

Download SCARED dataset. To obtain a link to the data and code release, sign the challenge rules and email them to [email protected]. You will receive a temporary link to download the data and code.

Follow MICCAI_challenge_preprocess to extract data. Put or link folders of the following cases into data/scared2019/dataset folder.

  • dataset_1_keyframe_1
  • dataset_2_keyframe_1
  • dataset_3_keyframe_1
  • dataset_6_keyframe_1
  • dataset_7_keyframe_1

data/scared2019/dataset should be as organized in the following structure.

├── data
│   ├── scared2019
│   │   ├── dataset
│   │   │   ├── dataset_1_keyframe_1
│   │   │   │   ├── data
│   │   │   │   │   ├── disparity
│   │   │   │   │   ├── frame_data
│   │   │   │   │   ├── left_finalpass
│   │   │   │   │   ├── reprojection_data
│   │   │   │   │   ├── ...
│   │   │   │   ├── ...
│   │   │   ├── dataset_2_keyframe_1
│   │   │   │   ├── ...
│   │   │   ├── dataset_3_keyframe_1
│   │   │   │   ├── ...
│   │   │   ├── dataset_6_keyframe_1
│   │   │   │   ├── ...
│   │   │   ├── dataset_7_keyframe_1
│   │   │   │   ├── ...
│   │   ├── ...

Preprocess the dataset with data/scared2019/preprocess.py. It scales and transforms the scene into a unit sphere, similar to NeuS. It also produces masks based on the validation area of depth maps (results saved in data/scared2019/dataset/*/data_processed). A *.pkl file is also generated in data/data_info/scared2019 to save all information about the dataset. Add --show to visualize results.

python data/scared2019/preprocess.py --dset_dir data/scared2019/dataset/dataset_1_keyframe_1  --skip_every 2  # --show
python data/scared2019/preprocess.py --dset_dir data/scared2019/dataset/dataset_2_keyframe_1  --skip_every 1  # --show
python data/scared2019/preprocess.py --dset_dir data/scared2019/dataset/dataset_3_keyframe_1  --skip_every 4 # --show
python data/scared2019/preprocess.py --dset_dir data/scared2019/dataset/dataset_6_keyframe_1  --skip_every 8  # --show
python data/scared2019/preprocess.py --dset_dir data/scared2019/dataset/dataset_7_keyframe_1  --skip_every 8  # --show

You can also download the preprocessed *.pkl files which are used in original paper.

Example preprocessing is shown below.

Visualization of SCARED preoprocessing

The overall data directory should be organized in the following structure.

├── data
│   ├── data_info
│   │   ├── scared2019
│   │   │   ├── dataset_1_keyframe_1.pkl
│   │   │   ├── dataset_2_keyframe_1.pkl
│   │   │   ├── dataset_3_keyframe_1.pkl
│   │   │   ├── dataset_6_keyframe_1.pkl
│   │   │   ├── dataset_7_keyframe_1.pkl
│   ├── scared2019
│   │   ├── dataset
│   │   │   ├── dataset_1_keyframe_1
│   │   │   │   ├── data_processed
│   │   │   │   │   ├── ...
│   │   │   │   ├── ...
│   │   │   ├── dataset_2_keyframe_1
│   │   │   │   ├── ...
│   │   │   ├── dataset_3_keyframe_1
│   │   │   │   ├── ...
│   │   │   ├── dataset_6_keyframe_1
│   │   │   │   ├── ...
│   │   │   ├── dataset_7_keyframe_1
│   │   │   │   ├── ...
│   │   ├── ...
│   ├── ...