Table 8.

Experiments that covered TinyML for ecology.

Study Reference	Year	Results and Implications
Alati, M. [33]	2022	Greenhouse temperature forecasting successfully implemented under extremely low power constraints of 0.17 mW.
Andrade, P. [32]	2022	Vehicle emission detection accuracy rate of 94% within a 1 ms inference latency constraint.
Alongi, F. [34]	2020	RMSE of 0.0255 on an MCU-embedded model of 512 kB memory.
Ogore, M. [35]	2021	94% detection accuracy of cholera on an offline device within latency, model size, and power constraints.
Trivedi, K. [38]	2021	88.3% detection accuracy with a 337 ms delay.
Falaschetti, L. [40]	2021	98.0% classification accuracy at 64.2 ms per image inference time (15.5 fps), with ~13 kB ROM used.
Du, P. [41]	2022	Successful prediction of plant growth and disease status within a 136 kB model, model usable in battery-powered camera systems.
Bruno, C. [42]	2021	Embedded model has a 72% accuracy in detecting and classifying gas samples.
Hayajineh, A. [47]	2024	Validation accuracy of 99%, results found that fewer LSTM structures resulted in faster inference and less memory usage.