Research Article

Nonnegative Matrix Factorizations Performing Object Detection and Localization

Table 3

Algorithm performances in terms of recall and precision when applied to Usps with factor ranks 𝑟 =80 and 𝑟 =220. Bold fonts indicate the highest values of precision and recall.

           𝑟 = 8 0
Method 𝑇 𝑃 𝐹 𝑃 R e c a l l P r e c i s i o n F-measure

NMF 2602 98 0.96 0.96 0.96
LNMF 2457 243 0.91 0.91 0.91
NMFsc 2615 850.970.970.97
DLPP 658 2042 0.24 0.24 0.24

           𝑟 = 2 2 0
Method 𝑇 𝑃 𝐹 𝑃 R e c a l l P r e c i s i o n F-measure

NMF 2602 980.960.960.96
LNMF 1708 1042 0.63 0.63 0.63
NMFsc 2603 970.960.960.96
DLPP 2195 505 0.81 0.81 0.81