Table 2.

Results for the multitask learning (MTL) network vs. the single task network for leaf counting task trained on the Ara2013 dataset.

Model	DiC	│DiC│	Agreement	MSE	R2
Single task	0.40 (1.09)	0.80 (0.84)	41	1.35	0.92
Multitask	0.22 (0.93)	0.67 (0.69)	45	0.93	0.95

All parameters are improved in the MTL model, with the mean squared error (MSE) showing an improvement of 40%.