Work program
The BioIntAkt project is composed of seven work packages, which are described as follows.
| Work package | Description |
|---|---|
| WP1 | At the beginning of the project, all requirements to achieve the project’s contributions are determined. The two primary use cases (“nature conservation in flower strips” and “oilseed rape agriculture”) are specifically taken into account, in order to ensure the generalizable nature of the technology to be developed. In addition to defining how the acoustic data will be collected, intended real-world deployment conditions are also considered in the definition of requirements to their mechanical robustness. |
| WP2 | The smart acoustic sensors for the project will be developed on the basis of reprogrammable hardware. This type of system design is both expandable and energy-efficient, so that even long-term measurement campaigns can be performed. Based on the requirements collected in WP1, the system design will be carried out in such a way that the execution of the AI models to be developed in WP4 is possible. In addition, the partners will be working on communication interfaces for wireless data exchange and evaluating the function of the systems in practice. |
| WP3 | To train the AI models to be developed in WP4, relevant indicator species are observed during various activities and annotated acoustic recordings will be created. The main goal of this work package is to create a catalog of the acoustic spectrum of various insects, which is not yet available worldwide. By means of an extensive measurement campaign, an unprecedented dataset of insect sounds will be created. Moreover, tools for semi-automated data recording, enhancements of the measurement environment with digitally controllable stimuli, or the conversion of recorded data into a file format that allows for simple processing, will be explored. |
| WP4 | Once the fundamental requirements for the identification of insect species from ecoacoustic recordings have been made in WPs 1-3, the AI model to perform the actual classification is designed next. For this purpose, a deep learning model will be developed and trained on extensive data sets. In addition to the classification of insects based on their emitted sounds, functions for the detection and removal of background noise from sound recordings, for the compensation of volume differences, and for the extraction of relevant features from the data will be developed. In addition, all the necessary technical preparatory work will be carried out to enable the model to be run in a cloud environment, where it will serve as the basis for decentralized audio data analysis using citizen science tools. |
| WP5 | A cloud backend and a mobile smartphone app for biodiversity factor analysis are being developed in this work package. Collected acoustic data will initially be processed in the cloud, thanks to the greater scalability and available computing power. The development of a smartphone app is crucial, and will enable interested citizens to make their own acoustic recordings using the microphone built into their smartphones, which can then be analyzed using the AI model. |
| WP6 | A data analysis platform for identifying trends and predictive models is being developed to spatially and temporally correlate the collected measurement data. This allows long-term predictions of biodiversity changes to be made, as well as catering to the identification of dependencies between biodiversity factors and environmental parameters. A dashboard makes it possible to visualize and model the spatial and temporal course of detected insect occurrences and to extrapolate their future distribution. The extensive database thus available allows targeted recommendations for action (e.g., the use of pesticides) to be made in an informed manner. |
| WP7 | The reliability and robustness of the acoustic data analysis will be evaluated by means of validation studies at various locations in agricultural areas and natural habitats. Based on the evaluation results, boundary conditions will be derived that govern how the measurement systems can be used in other geographic regions and for other types of animals, enabling their further usage even after the completion of the project. |