Using Audio IoT Sensors and RNNs to Monitor Honey Bee Colony Strength

Honey Bee Colony Monitoring via Audio IoT Sensors, Tensorgrams and RNNs

I explored how keeping the time dimension in modulation tensorgrams improves the monitoring of honey bee colony strength. By feeding this new representation into Convolutional Recurrent Deep Neural Networks trained on the UrBAN dataset, we achieved better accuracy and cross-hive generalizability than previous methods. Our results show that robust, accurate acoustic monitoring of hives is possible even in noisy, real-world conditions.