Table 2.

Classification accuracy in percent of the presented datasets using contrast-based windowing and state-of-the-art classification methods.

Method	Classifier	Balls	Beans
Intensity frames	ResNet-18	$81.37$	$\mathbf{100}$
Reconstructed frames [15]	ResNet-18	$\mathbf{92.37}$	$98.86$
HATS [18]	ResNet-18	$90.69$	$99.04$
Matrix-LSTM [19]	ResNet18-Ev2Vid	$89.06$	$99.44$
SPA [20]	SNN	$58.69$	$64.47$