Project description
The purpose of the project was to develop machine learning architecture for real time detection of fish in tanks using stereo cameras to track the swimming patterns of fish in 3D. A new machine learning architecture was developed, called StereoAttention. It was analysed on processing speed, complexity and accuracy. The fast algorithm for stereo detection was developed based on modern mechanisms. The proof of concept adds a solid foundation for further development.
Consortium
Coordinator:
Bjarne Kvæstad, SINTEF Ocean