Nonnegative Matrix Factorizations Performing Object Detection and Localization

Table 2

Algorithm performances in terms of 𝑟𝑒𝑐𝑎𝑙𝑙 and 𝑝𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 when applied to CarData with factor ranks 𝑟=20 and 𝑟=110. Bold fonts indicate the highest values of precision and recall.

𝑟=20
Method	𝑇𝑃	𝐹𝑃	Recall	Precision	F-measure
NMF	103	67	0.52	0.61	0.56
LNMF	92	78	0.46	0.54	0.5
NMFsc	106	64	0.53	0.62	0.57
DLPP	37	133	0.19	0.22	0.2
𝑟=110
Method	𝑇𝑃	𝐹𝑃	Recall	Precision	F-measure
NMF	112	58	0.56	0.66	0.61
LNMF	86	85	0.43	0.5	0.46
NMFsc	110	60	0.55	0.65	0.59
DLPP	21	93	0.11	0.18	0.13