TBD...
Clone the repo:
$ git clone https://github.com/nimiCurtis/bagtoolInstall the dependencies by:
$ pip install -r requirements.txt Build the package:
$ pip install -e . Important: This package is designed to work with bag files structured in accordance with the recording conventions established by the zion_zed_ros_interface. Please ensure compatibility with this structure for optimal performance. for more info on the structure use the given link
The main tool of this package is the process tool, which process ROS bag files from a given folder path and saving the data in useful formats and structures for future works.
- From the CLI :
bagtool process [-h] [-b BATCH] [-f FOLDER] [-d DST] [-n NAME]
[--no_raw]
optional arguments:
-h, --help show this help message and exit
-b BATCH, --batch BATCH
path to a bag batch folder
-f FOLDER, --folder FOLDER
path to a bag folder consisting bag batches
-d DST, --dst DST path to a dataset destination folder
-n NAME, --name NAME a custom name for the datafolder
--no_raw disable raw data saveing
- And From a custom script using the bagtool.process module
Assuming you have a folder with a bag batch folder (see the important note) at <some_path>/bag folder. And you'd like to set your entire dataset on <some_path>/dataset.
Using CLI:
bagtool process -f <some_path>/bag --dst <some_path>/dataset
Using custom script:
from bagtool.process.process import BagProcess as bp
source = '<some_path>/bag'
dest = '<some_path>/dataset'
bp.process_folder(source,dest)(Recomended) Using custom script + .yaml config:
We can a extract the dataset from a recordings database using custom config file.
Such a config would looks as follow:
bags_folder: "/home/roblab20/catkin_ws/src/zion_ros/zion_zed_ros_interface/bag"
destination_folder: "/home/roblab20/dev/bagtool/dataset"
save_raw: true # I dont see a reason why to set it to false
demonstrator_name: # demonstrator/robot name
aligned_topics: ["odom", "rgb", "depth", <topic key name>] # pick topics to be aligned
sync_rate: 20 # alignment rate, if null we will
# use the min frequency, from the aligned topics frequency
save_vid: true # saving video
## more params .. Process file would looks as follow:
import yaml
from bagtool.process.process import BagProcess as bp
def main():
config_path = 'process_bag_config.yaml'
with open(config_path, 'r') as file:
process_config = yaml.safe_load(file)
bp.process_folder(config=process_config)
if __name__ == "__main__":
main()Dataset structure:
The resulted dataset from the example above will be as follow:
├── dataset
│ ├── <name_of_bag_batch1>
| | ├── raw_data
│ │ ├── raw_<topic1>.h5
│ │ ├── ...
│ │ └── raw_<topicN>.h5
| |
5B81
├── visual_data
| | ├── depth
│ │ ├── 0.jpg
│ │ ├── ...
│ │ └── T.jpg
| | └── rgb
│ │ ├── ...
│ │ └── T.jpg
│ │ ├── metadata.json
│ │ ├── robot_traj_data.json
│ │ ├── target_traj_data.json (when using the object detection topic)
│ │ └── traj_sample.mp4
│ ...
└── └── <name_of_bag_batchN>
├── raw_data
├── raw_<topic1>.h5
├── ...
└── raw_<topicN>.h5
├── visual_data
├── depth
├── 0.jpg
├── ...
└── T.jpg
└── rgb
├── ...
└── T.jpg
├── metadata.json
├── robot_traj_data.json
├── target_traj_data.json (when using the object detection topic)
└── traj_sample.mp4
Another tool provided is a quick & simple command line interface for rosbag compressing of entire batches of bags and folder of batches.
- (Recomended) From the CLI :
usage: bagtool compress [-h] [-b BATCH] [-f FOLDER]
optional arguments:
-h, --help show this help message and exit
-b BATCH, --batch BATCH
path to a bag batch folder
-f FOLDER, --folder FOLDER
path to a bag folder consisting bag batches
By priority Top-Down:
- Improve efficiency and modularity
- Handling with 32bit depth images ?
- Check use of the compressed Image sensor msgs
- Add arrow to the traj animation indicating the yaw direction ?