Installation

Backround: Data Source

The data in this paper is sourced from the experiments described and conducted in the paper Plume-Tracking Behavior of Flying Drosophila Emerges from a Set of Distinct Sensory-Motor Reflexes (F. van Breugel et al., 2014). Three data sets are used, corresponding to measurements of flies flying in an artificial wind of 30 cm/s, 40 cm/s, and 60 cm/s. Each data set is itself composed of three subsets of data: flight trajectory data, body orientation data, and a key table to correlate the two for a given fly. All of this data is may be found at my Dropbox with the following directory tree:

ExperimentalData
├── 30cms
│   ├── flight_trajectories_3d_HCS_odor_horizon_matched.h5
│   ├── body_orientations_HCS_odor_horizon_matched.h5
│   └── body_trajec_matches.h5
├── 40cms
│   ├── flight_trajectories_3d_HCS_odor_horizon_matched.h5
│   ├── body_orientations_HCS_odor_horizon_matched.h5
│   └── body_trajec_matches.h5
└── 60cms
    ├── flight_trajectories_3d_HCS_odor_horizon_matched.h5
    ├── body_orientations_HCS_odor_horizon_matched.h5
    └── body_trajec_matches.h5

1. Clone this repo:

git clone https://github.com/nehalsinghmangat/drosophila_body_orientation_predictor.git

2. Download data:

Next, download ExperimentalData from my personal dropbox here and clobber the existing ExperimentalData directory in the repo.

3. Build docker image:

If docker is not already installed on your system, please install using the official Docker documentation. Then navigate inside the drosophila_body_orientation_predictor repo and run the following command:

sudo docker build -t body_dros_venv .

This creates the docker image body_dros_venv, which you can view by running sudo docker images ls.

4. Run docker container

The docker container contains an instance of jupyter-lab whose kernel contains all dependencies required for the analysis. To create the container and run it, run the following command inside the repo:

sudo docker run -it -v .:/home --rm -p 8888:8888 body_dros_venv

This will start the notebook. The -it flag allows the user to interact with the container. The -v flag mounts the drosophila_body_orientation_predictor host directory to the /home client directory; writing to one will write to the other, allowing the user to persist any data generated during the analysis. The -p flag maps the container port 8888 to the host port 8888 (this port must be available on your machine). To access the notebook, simply copy and paste the local URL generated in the container into your host's browser (the URL should begin with http://127.0.0.1:8888.) The --rm flag is discussed in step 6.

5. Add MOSEK license

The jovyan directory is the default user directory for the jupyter-lab notebook. Because we use the MOSEK solver in cvxpy, the user must add a valid mosek.lic license to jovyan/mosek. If not, the user will be unable to correct the heading angles in the fly data.

6. Exiting safely

Because we ran the docker container with the --rm flag, upon exiting the container, the docker engine will automatically delete the container and its image. This will help keep the host machine's memory clean. If one wishes to persist the changes made in the docker container (for development purposes, say) then run step 4 without the --rm flag. This time, when you exit the container, it is stopped. To restart it and resume your session, run

sudo docker start --attach -i container_id

Note: you can run sudo docker ps -a to find the container id.