bird-behavior

A tool for classifying bird behaviors from sensor data.

Installation

The minimum required Python version is 3.10. If you don't have a newer version of Python installed, using conda or miniconda is the cleanest way to set it up. To install conda or miniconda, follow these instructions.

The following commands are required when using conda:

conda create --name bird python=3.10 -y  # Create a virtual environment
conda activate bird                      # Activate the environment
conda deactivate                         # Deactivate the environment
conda remove --name bird --all           # Remove the virtual environment

Clone the repository and install the requirements:

git clone https://github.com/fkariminejadasl/bird-behavior.git
cd bird-behavior
# Ensure the conda environment is activated
# Install the software
pip install -e .
# Install additional requirements for notebooks (if needed)
pip install -e .[notebook]

Usage

General Users

If you do not have access to our database, you can use your own data in the format described below to classify bird behaviors. You need to download a pre-trained model, which can be found here.

input format

The data.csv file is a simple text file where each row contains the following fields, separated by commas:

• device_id
• date_time (format: year-month-day hour:minute:second)
• index (can be 0 or increasing values)
• label (should be -1 indicating no label)
• IMU x, y, z acceleration
• GPS 2D speed . Here is an example (also available in data/data.csv)`:

537,2013-06-06 20:34:11,0,-1,0.36737926,0.03602193,1.59495666,9.84760679
537,2013-06-06 20:34:11,0,-1,0.38796516,0.11902897,1.41765169,9.84760679

Command to Run

python scripts/classify_birds.py --data_file /your_path/manouvre.csv  --model_checkpoint /your_path/model.pth --save_path /your_path/result
# Alternatively, with default values in the config file:
python scripts/classify_birds.py  configs/classification.yaml --data_file /your_path/data.csv 
# Or by specifying everything in the config file:
python scripts/classify_birds.py  configs/classification.yaml

IBED Users

Use the notebooks/classify_birds.ipynb or follow the instruction below.

Input Format

The input.csv file is a simple text file where each row contains the following fields, separated by commas:

• device_id
• start_time (format: year-month-day hour:minute:second)
• end_time (format: year-month-day hour:minute:second)

Here is an example (also available in data/input.csv):

541,2012-05-17 00:00:59,2012-05-17 00:00:59
805,2014-06-05 11:16:27,2014-06-05 11:17:27

command to run

python scripts/classify_birds.py --input_file /your_path/input.csv --username your_user --password your_pass  --model_checkpoint /your_path/45_best.pth --save_path /your_path/result
# Alternatively, with default values in the config file:
python scripts/classify_birds.py  configs/classification.yaml --input_file /your_path/input.csv --username your_user --password your_pass 
# Or by specifying everything in the config file:
python scripts/classify_birds.py  configs/classification.yaml