Internship: Training of embedded neural networks for bird song detection

Research Internship Project
EZBird – Training of embedded neural networks for bird song detection


Context
This internship subject takes place in the eBRAIN group of LEAT laboratory that works on Embedded
Bio-inspiRed Artificial Intelligence and Neuromorphic architectures.
More specifically, it takes part a collaboration with CERN in Geneva in order to design, develop and
deploy a wireless sensor network to monitor the biodiversity on the campus of CERN in Switzerland.
The goal being to design sensor nodes endowed with embedded AI [1] to automatically and
autonomously detect the presence of birds and classify the species according to their song [2,4].
The electronic part of the nodes has already been designed and validated. They are composed of a
microcontroler unit (MCU), a battery and a solar panel, and a set of sensors monitoring the
environmental conditions [3].
Each node has also a microphone specifically used to detect birds.
The main goal of this internship is to work on the AI part in order to i) record new data on site in
Geneva, ii) improve the performance of classification of the deep learning model, iii) and optimize
the neural network (in terms of memory footprint, latency and accuracy) in order to be integrated in
the sensor nodes before the final deployment on site, and iv) to analyze the data collected from 50
sensors over several weeks.


Project mission
The project mission will be organized in several periods:

  • bibliography on bird song detection on the Edge
  • collect audio data to build an augmented dataset
  • train and optimize a deep learning both in terms of performance, and memory footprint
  • integrate the model into the sensor nodes thanks to the QUALIA Sw framework
  • analyze the data collected over time
  • write a publication on this work in conference or journal

A part of the internship will thus take place in Geneva.


References
[1] Quantization and deployment of deep neural networks on microcontrollers, PE Novac, G Boukli Hacene, A. Pegatoquet, B Miramond, V Gripon, Sensors 21 (9), 2984, 2021
[2] Bird@Edge: Edge AI system for recognizing bird species in audio recordings to support real-time biodiversity monitoring, J Höchst, H Bellafkir, P Lampe, M Vogelbacher, Networked Systems, 2022 – Springer
[3] Nature 4.0: a networked sensor system for integrated biodiversity monitoring, D Zeuss, L Bald, J Gottwald, M Becker, 2024 – Wiley Online Library
[4] A survey on deep learning based forest environment sound classification at the edge, D Meedeniya, I. Ariyarathne, M Bandara, ACM Computing, 2023


Practical information
Location: LEAT Lab / SophiaTech Campus, Sophia Antipolis and CERN in Geneva.
Duration: 6 months from March 2025
Profile: Machine learning, Data science, Python programming, embedded programming
Research keywords: Embedded systems, Edge AI, Digital signal processing


Contact and supervision
Benoît Miramond, Pierre-Emmanuel Novac
LEAT Lab – University Cote d’Azur / CNRS
Polytech Nice Sophia
04.89.15.44.39. / firstname.name@univ-cotedazur.fr

Leave a Reply

Votre adresse e-mail ne sera pas publiée. Les champs obligatoires sont indiqués avec *