deepecohab 0.5.0

EcoHab with some machine learning

DeepEcoHab: fast and intuitive data analysis platform for your EcoHab experiments

DeepEcoHab is an analytics platform build for preprocessing, analysis and visualization of data acquired in the DeepEcoHab.

Our backend is built on Polars - Extremely fast Query Engine for DataFrames, written in Rust and frontend utilizes Plotly Dash which allows for system independent operation - running the app in your Chromium based browser - providing an interactive, high quality and responsive visualization of experiments regardless of their length.

Installation

We keep DeepEcoHab lean to ensure easy integration and fast installation.

Existing Environments:

If your environment is already running python>=3.10, run: pip install deepecohab

New Installations: If you are starting from scratch, please follow our guide below:

In the spirit of open-source we suggest usage of uv.

To install uv copy-paste the command below:

Windows: powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Linux/MacOS: $ curl -LsSf https://astral.sh/uv/install.sh | sh

To install DeepEcoHab please run the following commands line by line in the terminal:

Turn slashes the other way for Linux and MacOS

cd where\you\want\venv
uv venv
.venv\Scripts\activate
uv pip install deepecohab

We recommend using VSCode with the Jupter extension to run the example notebooks provided in the repository.

Example data

We provide 3 example datasets that reflect 3 main possibilites for an EcoHab layout.

• example_notebook for a vanilla 4 cage, 8 antenna setup.
• example_notebook_custom_layout for a custom layout that can be user defined in the config.toml of the created project.
• example_notebook_field for a field EcoHab layout.

Dashboard

The dashboard contains visualization of the experiment analysis results. It is divided into two tabs: main dashboard tab and a tab for comparisons (when the user wants to compare same plot in different days/phases etc.) and 3 sections:

1. Social hierarchy
2. Activity
3. Sociability

All providing multiple plots controlled via the settings block located on top.

Data structure:

The data is stored in parquet format - an open-source, column-oriented data storage format which allows extremely fast read/write operations of large dataframes.

To get the list of available keys simply call: deepecohab.df_registry.list_available() similarily deepecohab.plot_registry.list_available() can be called to obtain the list of currently available visualizations.

Roadmap

1. Full web-app style GUI, deployable via a docker container.
2. Group analysis - combined analysis of multiple cohort, comparing different groups of cohorts.
3. Pose estimation based analysis of animal interactions and more detailed social structure analysis.