/deepFEPE_data

# Data processing for pose estimation tasks

Primary LanguagePythonMIT LicenseMIT

Data processor

Data processor for pose estimation or visual odometry tasks.

KITTI dataset

Raw data structure

  • Download raw data from here.
  • Download odometry data (color) from here.
  • Copy the ground truth poses from deepFEPE/datasets/kitti_gt_poses.
`-- KITTI (raw data, odometry sequences, GT poses)
|   |-- raw
|   |   |-- 2011_09_26_drive_0020_sync
|   |   |   |-- image_00/
|   |   |   `-- ...
|   |   |-- ...
|   |   `-- 2011_09_28_drive_0001_sync
|   |   |   |-- image_00/
|   |   |   `-- ...
|   |   |-- ...
|   |   `-- 2011_09_29_drive_0004_sync
|   |   |   |-- image_00/
|   |   |   `-- ...
|   |   |-- ...
|   |   `-- 2011_09_30_drive_0016_sync
|   |   |   |-- image_00/
|   |   |   `-- ...
|   |   |-- ...
|   |   `-- 2011_10_03_drive_0027_sync
|   |   |   |-- image_00/
|   |   |   `-- ...
|   |-- sequences
|   |   |-- 00/
|   |   |-- ...
|   |   |-- 10/
|   |-- poses
|   |   |-- 00.txt
|   |   |-- ...
|   |   |-- 10.txt
`   `   `

Run command

WE ARE NOT FILTERING STATIC FRAMES FOR THE ODO DATASET! Set --with_pose --with_X --with_sift to decide whether to dump pose files, rectified lidar points, and SIFT kps/des and corres.

  • Specify your dataset_dir and dump_root
python dump_tools/dump_data.py --dump --dataset_dir /media/yoyee/Big_re/kitti/data_odometry_color/dataset/ \
--dump_root /media/yoyee/Big_re/kitti/kitti_dump/odo_corr_dump_siftIdx_npy_delta1_test_0714 \
--with_pose --with_sift \
--img_height 376 --img_width 1241 --dataloader_name kitti_seq_loader --cam_id '02'

ApolloScape dataset

Raw data structure

  • Download raw data (Training data, Road11.tar.gz) from here or use the following script.
python dump_tools/apollo/download.py -h
python dump_tools/apollo/download.py --dataset_dir /media/yoyee/Big_re/apollo/train_seq_1 --if_download
# change the name to Road11.tar.gz
tar zxf Road11.tar.gz

Run command

  • Specify your dataset_dir and dump_root
python dump_tools/dump_data.py  --dump --dataset_dir /media/yoyee/Big_re/apollo/train_seq_1/  --dataloader_name  apollo_train_loader  --with_pose    --with_sift --dump_root /media/yoyee/Big_re/apollo/apollo_dump/train_seq_1/  --cam_id 5  --img_height 2710 --img_width  3384

EuRoC dataset (not tested)

Raw data

python dump_tools/euroc/download.py --dataset_dir /media/yoyee/Big_re/euroc/train_seq_1 --if_download --if_untar

Process data

  • Match time stamps of rgb images, depth images and poses
python dump_tools/euroc/process_poses_euroc.py --dataset_dir /data/euroc/test2/ --dataset euroc

Run command

python dump_tools/dump_data.py  --dump --dataset_dir  /data/euroc/raw_sequence --with_pose     --with_sift --dump_root /data/euroc/euroc_dump/slam_seq_v1   --cam_id 00 --img_height 480 --img_width 752 --dataloader_name euroc_seq_loader

TUM dataset (not tested)

Raw data

python dump_tools/tum/download.py

Run command

python dump_tools/dump_data.py  --dump --dataset_dir /data/tum/raw_sequences  --with_pose     --with_sift --dump_root /data/tum/tum_dump/slam_seq_v1 --with_X   --cam_id 00  --img_height 480 --img_width 640  --dataloader_name tum_seq_loader

Visualize dataset

Refer to https://github.com/eric-yyjau/kitti_instance_RGBD_utils for some code snippets.

Citations

Please cite the following papers.

  • DeepFEPE

Credits

This implementation is developed by You-Yi Jau and Rui Zhu. Please contact You-Yi for any problems.

License

DeepFEPE is released under the MIT License.