|
Author | Dataset | Accuracy | Summary |
|
Emeršič et al. [60] | NA | 30 | It used handcrafted feature extraction methods such as LBP, POEM, and CNN to obtain the ear identification |
Tian et al. [21] | AMI, WPUT, IITD, and UERC | 70.58, 67.01, 81.98, and 57.75 | This system used deep CNN to perform ear recognition. There were occlusions like no earrings, headsets, or similar occlusions |
Raveane et al. [64] | NA | 98 | This system used variable conditions due to the odd shape human ear and changing lighting conditions |
Zhang and Mu [65] | UND and UBEAR | 100 and 98.22 | This system contained large occlusions, scale, and pose variation |
Kohlakala and Coetzer [66] | AMI and IIT-Delhi | 99.2 and 96.06 | It is used to classify ears either in the foreground or background of the image. The binary contour image applied the matching for feature extraction, and this was performed by implementing Euclidean distance measure, which had a ranking to verify for authentication |
Tomczyk and Szczepaniak [67] | NA | NA | It shows the published experimental results that the approach did the rotation equivalence property to detect rotated structures |
Alshazly et al. [68] | Three ear datasets but not stated | 22 | The paper took seven performing handcrafted descriptors to extract the discriminating ear image. Then took the extracted ear and trained it using SVM to learn a suitable model |
Alkababji and Mohammed [69] | NA | 97.8 | It used the PCA and a genetic algorithm for feature reduction and selection |
Jamil et al. [70] | Very underexposed or overexposed database | 97 | This work was the first to test the performance of CNN on very underexposed or overexposed images |
Hansley et al. [71] | UERC challenge | NA | This was performed using handcrafted descriptors, which were fused to improve recognition |
|