Impact of Occlusion Masks on Gender Classification from Iris Texture

Table 1

Summary of works on gender classification from iris.

Citation	Year	Spectrum	Image type	Occlusion masks	Best accuracy (%)
[1]	2015	NIR	Normalized	✓	91.33
[35]	2015	NIR	Normalized	✗	89.74
[30]	2016	NIR	Normalized	✓	91.00
[9]	2016	NIR	Both	✗	69 Normalized, 85.7 periocular
[15]	2017	NIR	Periocular	✗	83.17
[10]	2017	NIR/VIS	Periocular	✗	89.59
[16]	2017	VIS	Periocular	✗	90.20
[31]	2017	NIR	Normalized	✓	84.66
[14]	2017	NIR	Both	✓	66 Normalized, 80 periocular
[17]	2018	VIS	Periocular	✗	90.00
[18]	2018	NIR	Periocular	✗	87.26
[11]	2018	NIR	Periocular	✗	85.93
[12]	2018	NIR	Periocular	✗	85.90
[19]	2019	NIR	Periocular	✗	85.40
[32]	2019	NIR	Normalized	✓	95.45
[4]	2019	NIR	Both	✓	63.40 Normalized, 80.80 periocular
[20]	2019	NIR	Periocular	✗	94.63
[21]	2019	NIR/VIS	Periocular	✗	86.89
[22]	2019	NIR/VIS	Periocular	✗	91.90
[23]	2019	NIR	Periocular	✗	91.90
[36]	2019	NIR	Normalized	✓	94.66
[24]	2019	VIS	Periocular	✗	90.15
[25]	2019	NIR	Both	✗	66.67 Normalized, 93.34 face
[26]	2020	VIS	Periocular	✗	92.00
[27]	2020	NIR/VIS	Periocular	✗	81.59
[37]	2020	NIR	Normalized	✗	96.00
[38]	2021	VIS	Periocular	✗	75.10
[39]	2021	NIR/VIS	Periocular	✗	75.72
[40]	2023	NIR	Both	✗	98.92 Normalized, 73.96 periocular
[41]	2023	VIS	Periocular	✗	98.71

Works that do not use the NIR spectrum have not been thoroughly included. Image type “Both” means experiments were performed with normalized and periocular images.