Wildlife Monitoring on the Edge: A Performance Evaluation of Embedded Neural Networks on Microcontrollers for Animal Behavior Classification

Abstract

Monitoring animals’ behavior living in wild or semi-wild environments is a very interesting subject for biologists who work with them. The difficulty and cost of implanting electronic devices in this kind of animals suggest that these devices must be robust and have low power consumption to increase their battery life as much as possible. Designing a custom smart device that can detect multiple animal behaviors and that meets the mentioned restrictions presents a major challenge that is addressed in this work. We propose an edge-computing solution, which embeds an ANN in a microcontroller that collects data from an IMU sensor to detect three different horse gaits. All the computation is performed in the microcontroller to reduce the amount of data transmitted via wireless radio, since sending information is one of the most power-consuming tasks in this type of devices. Multiples ANNs were implemented and deployed in different microcontroller architectures in order to find the best balance between energy consumption and computing performance. The results show that the embedded networks obtain up to 97.96% ± 1.42% accuracy, achieving an energy efficiency of 450 Mops/s/watt.

TinyCowNet: Memory- and Power-Minimized RNNs Implementable on Tiny Edge Devices for Lifelong Cow Behavior Distribution Estimation

“…[27] per inference. By contrast, the highestaccuracy algorithm discussed in Section II at 98.3% has a 185.8 mJ energy consumption per inference [25]. This energy consumption is drastically higher than ours.…”

“…This MLP was implemented on a Nucleo MCU (ST Microelectronics, Spa, Agrate Brianza, Italy) using the Fast Artificial Neural Network library [31]. Later work shows an accuracy of 98.3% on the same three behaviors with a 163kB three-layer MLP floating-point model on a similar MCU, consuming 185.76mJ per inference [25].…”

“…Fig. 1 shows a cow behavior distribution estimating RNN model created in this work with an 860k× lower energy consumption than the highest-consuming estimation algorithm [25] with 8.4% higher accuracy than the lowest energyconsuming ones [27].…”

An overview of remote monitoring methods in biodiversity conservation

“…Costing Nature is another easy-to-use rapid and reliable web-based technique used for screening protected areas, land use and land cover (LULC), trends of habitation, biodiversity assessment, and possible future threats using global database. It has been used for testing ES for timber, fuel wood, grazing/fodder, and non-wood forest products (Thessen 2016 ; Dominguez-Morales et al 2021 ; Neugarten et al 2018 ).…”

Identification of Factors Affecting Environmental Contamination Represented by Post-Hatching Eggshells of a Common Colonial Waterbird with Usage of Artificial Neural Networks

“…ANNs are more and more frequently applied to the analysis of the results of experimental research conducted in many research programs [ 23 , 24 , 25 , 26 , 27 ]. Neural modeling can be effectively used to solve classification and regression problems in biological sciences, including prediction [ 28 , 29 ].…”