Holistic and component plant phenotyping using temporal image sequence

Table 5 Performance summary of algorithm 1 on UNL-CPPD dataset (Naming convention for plant sequence is: Plant_ID-Genotype ID [1])

Plant sequence	Dataset	No. leaves	Detected leaves	False leaves	Accuracy
\(Plant\_{001}-9\)	CPPD-I	116	93	1	0.79
\(Plant\_{001}-9\)	CPPD-II	168	157	5	0.83
\(Plant\_{006}-25\)	CPPD-I	138	136	0	0.98
\(Plant\_{006}-25\)	CPPD-II	205	188	5	0.91
\(Plant\_{008}-19\)	CPPD-I	142	140	0	0.98
\(Plant\_{008}-19\)	CPPD-II	210	200	9	0.86
\(Plant\_{016}-20^+\)	CPPD-I	103	86	0	0.83
\(Plant\_{016}-20^+\)	CPPD-II	141	129	0	0.88
\(Plant\_{023}-1\)	CPPD-I	113	101	0	0.89
\(Plant\_{023}-1\)	CPPD-II	154	135	8	0.83
\(Plant\_{045}-1\)	CPPD-I	122	120	3	0.96
\(Plant\_{045}-1\)	CPPD-II	177	170	6	0.93
\(Plant\_{047}-25\)	CPPD-I	148	142	2	0.94
\(Plant\_{047}-25\)	CPPD-II	212	196	5	0.88
\(Plant\_{063}-32^\dagger\)	CPPD-I	149	138	0	0.93
\(Plant\_{063}-32^\dagger\)	CPPD-II	214	174	18	0.72
\(Plant\_{070}-11\)	CPPD-I	125	111	0	0.89
\(Plant\_{070}-11\)	CPPD-II	177	148	5	0.83
\(Plant\_{071}-8\)	CPPD-I	141	131	0	0.93
\(Plant\_{071}-8\)	CPPD-II	199	163	7	0.77
\(Plant\_{076}-24\)	CPPD-I	135	126	2	0.92
\(Plant\_{076}-24\)	CPPD-II	191	152	2	0.78
\(Plant\_{104}-24^\ddagger\)	CPPD-I	144	140	0	0.97
\(Plant\_{104}-24^\ddagger\)	CPPD-II	186	185	0	0.96
\(Plant\_{191}-28\)*	CPPD-I	137	111	0	0.96
\(Plant\_{191}-28\)*	CPPD-II	178	151	7	0.81
Average	CPPD-I	132	123	< 1	0.92
Average	CPPD-II	186	165	\(\approx\) 6	0.85

* Plant sequence used to demonstrate inaccuracy in leaf detection due to self-occlusion and leaf crossover
⁺Plant-level accuracy for UNL-CPPD-II is higher than that of UNL-CPPD-I
^†Plant-level accuracy for UNL-CPPD-II is lower than that of UNL-CPPD-I
^‡Plant-level accuracy remains fairly similar for both UNL-CPPD-I and UNL-CPPD-II

ISSN: 1746-4811