Context & Issue
Avian mortality due to collision with wind turbine blades is very high. To reduce or even eliminate these accidents, rapid and accurate detection of the birds is necessary in order to take measures to scare them away or to stop the machines.
Summary
Project duration
4
months
Team
1
person
Aims
- Real-time detection of flying fauna
- Estimation of species from video streams
- Estimation of the distance of the birds
Realisation
- Dimensioning of the hardware infrastructure
- Modelling (object detection with an EfficientDet)
- Production launch of the models (TFX)
Technologies used
- Azure
- Open CV
- Python
- TensorFlow Extended
- Nvidia
- Tensor RT
