A tool for automatically segmenting videos based on depth data. Used for region-of-interest segmentation and selection for videos of neonatal patients in the NICU. Developed as part of a project by the Carleton University - Biomedical Informatics Co-laboratory (CU-BIC).
Features include:
- Head and torso region-of-interest segmentation
- Perspective transformation of recorded depth video
- Compatibility with rosbag (.bag) files recorded by Intel RealSense RGB-D cameras
- Combining and encoding rosbag files to mj2 archival video files.
The tools were tested and confirmed to run with Python 3.8.10.
The prefered way to install the python dependencies is through a Conda environment:
conda env update --file environement.yml
You can also install the python dependencies for the tools from the 'requirements.txt' file using pip instead:
python -m pip install -r requirements.txt
The scripts found at 'bagmerge/rosbag_merge.py' and 'bagmerge/realsense_rosbag_merge.py' require an active ROS Noetic environment to run properly. Instructions to install on Ubuntu (or WSL) can be found here.
The scripts to encode rosbag files as mp4 or mj2 videos were built and tested in MATLAB R2019a.
WIP
WIP
As of right now the development environment is the same as the one used to run the scripts.
WIP
[WIP]