Skip to content

Latest commit

 

History

History
91 lines (69 loc) · 5.68 KB

README.md

File metadata and controls

91 lines (69 loc) · 5.68 KB
Raw

Analysing antennal positioning during the 🐝 dance

Paper: Hadjitofi, A. & Webb, B. (2024). Dynamic antennal positioning allows honeybee followers to decode the dance. Current Biology, https://doi.org/10.1016/j.cub.2024.02.045

Data: Hadjitofi, A. & Webb, B. (2024). Honeybee antennal positioning data when following dances. figshare. Dataset. https://doi.org/10.6084/m9.figshare.24715977

Example video (Video S1):

Video thumbnail

Running the assimilation model 🧠

The assimilation model can be run for different follower - dancer orientations in simulation (perfect and noisy) as well as on the antennal positioning data from real bees (labelled data).

[17:56:23] 🚀 dance2vec $ python simulate_antenna.py --help
usage: simulate_antenna.py [-h] [--cx_noise CX_NOISE] [--cx_cpu4_memory_gain CX_CPU4_MEMORY_GAIN] [--recruit_antenna_flow_method {midpoint,left_only,right_only,single_antennae}] [--recruit_dont_clip_antenna_angles]
                           [--experiment_name EXPERIMENT_NAME] [--cx_model {CXRecruitHoloBee,CXRecruitHoloFly}] [--recruit_angle_to_gravity RECRUIT_ANGLE_TO_GRAVITY] [--dancer_angle_to_gravity DANCER_ANGLE_TO_GRAVITY]
                           [--max_antenna_pos MAX_ANTENNA_POS] [--add_noise] [--simulation_time SIMULATION_TIME] [--dt DT] [--antennal_positioning_dataset ANTENNAL_POSITIONING_DATASET] [--bee_id BEE_ID]
                           [--remove_antennae_input] [--seed SEED]

Configuration for experiments. Inspired by https://github.com/nuric/pix2rule

optional arguments:
  -h, --help            show this help message and exit

CX options:

  --cx_noise CX_NOISE   Level of noise to add to layer outputs of CX. (default: 0.0)
  --cx_cpu4_memory_gain CX_CPU4_MEMORY_GAIN
                        Rate of memory accumulation. (default: 0.005)

Recruit options:

  --recruit_antenna_flow_method {midpoint,left_only,right_only,single_antennae}
                        Type of CX model to use in experiments. (default: midpoint)
  --recruit_dont_clip_antenna_angles
                        Whether or not to clip antennae angles if they exceed the maximum antennae position specified. (default: False)

Simulation options:

  --experiment_name EXPERIMENT_NAME
                        Experiment name, default current datetime. (default: 20240207-1756)
  --cx_model {CXRecruitHoloBee,CXRecruitHoloFly}
                        Type of CX model to use in experiments. (default: CXRecruitHoloFly)
  --recruit_angle_to_gravity RECRUIT_ANGLE_TO_GRAVITY
                        Recruit's angle to gravity (degrees), going ccw from 0 N to 90 W, +- 180 S, -90 E. (default: None)
  --dancer_angle_to_gravity DANCER_ANGLE_TO_GRAVITY
                        Dancer's angle to gravity (degrees), going ccw from 0 N to 90 W, +- 180 S, -90 E. (default: None)
  --max_antenna_pos MAX_ANTENNA_POS
                        Max +ve valid angle of antenna relative to bees midline (degrees). Used if running prefect simulation. (default: None)
  --add_noise           Add noise to any simulated antennae positions. (default: False)
  --simulation_time SIMULATION_TIME
                        Maximum simulation time (in seconds) per trial. (default: 3)
  --dt DT               Time resolution to sample. Default records at base time (samples once per second) (default: 0.01)
  --antennal_positioning_dataset ANTENNAL_POSITIONING_DATASET
                        Path to real antennal positioning dataset. Defaults to None (i.e. no use). If provided, will feed this data to the model instead of simulating positions. (default: None)
  --bee_id BEE_ID       Index of bee to monitor if antennal positioning dataset is given. If not specified, a random bee id will be visualised. (default: None)
  --remove_antennae_input
                        Don't use antenna info. Only available when using real antennal positions. (default: False)
  --seed SEED           Seed to set. (default: 1)

Analysing the data 📈

Two datasets will need to be built in order to analyse (1) the raw antennal positions (from labelled data) and (2) the errors of the resulting flight vectors from the assimilation model with the corresponding antennal positioning data.

(.venv) [17:31:50] 🚀 roslin-2022-analysis $ python build_dataset.py --help
usage: build_dataset.py [-h] [--build_errors_ds] [--data_path DATA_PATH] [--file_out FILE_OUT] [--nbins NBINS] [--antenna_len_str {default,mid_only,full_only}]

Build antennal positioning dataset.

optional arguments:
  -h, --help            show this help message and exit
  --build_errors_ds     Build error dataset instead of antennae dataset.
  --data_path DATA_PATH
                        Name of raw data folder. Defaults to 'Cropped-Anna Hadjitofi-2022-11-01'. For building errors dataset, this should be the path to the folder containing the experiments.
  --file_out FILE_OUT   Resulting name (and path) to save dataset(s), without extension. Defaults to current date.
  --nbins NBINS         Number of bins to use for binning angles to dancer. Defaults to 180.
  --antenna_len_str {default,mid_only,full_only}
                        Calculate antenna angle using mid length / bend or full length. The default uses base to tip if available and midpoint as fallback.

Built With 🔨

  • Matplotlib - main plotting library
  • seaborn - helper plotting library for some charts
  • NumPy - main numerical library for data vectorisation
  • Pandas - helper data manipulation library
  • OpenCV - video loading and extraction library