Table 9.

Experiments that covered TinyML usage in vehicular detection.

Study Reference	Year	Results and Implications
Falaschetti, L. [36]	2024	Detection precision of pedestrians of ~77% on a model compressed to ~33% of its original size.
Zhang, S. [37]	2020	Bus passengers detected with high accuracy on a compressed model.
Andrade, P. [43]	2021	Speed bumps and potholes detected with F1 score of ~0.76; model deployed on an Arduino Nano.
Alajlan, N. [39]	2023	Quantized models were assessed for accuracy and model size. CNN model had greatest compression (0.05 MB), while the DRQ MobileNet-V2 model had the highest accuracy at 0.9964.