The Unconstrained Ear Recognition Challenge 2019

Emeršič, Žiga; Park, H. J.; Nam, Gi Pyo; Kim, I.-J.; Sangodkar, S.G.; Kacar, Umit; Kırcı, Mürvet; Li, Yuan; Yuan, J.; Zhao, Haoxiang; Lü, Fei; Mao, Junying; Zhang, X.; Yaman, D.; Eyiokur, Fevziye Irem; Özler, Kadir Bulut; Ekenel, Hazım Kemal; Chowdhury, Debbrota Paul; Bakshi, Sambit; K., P.; Majhi, Banshidhar; Harish, B. S.; Peer, Peter; Gutfeter, Weronika; Khiarak, Jalil Nourmohammadi; Pacut, Andrzej; Hansley, Earnest E.; Segundo, Maurício Pamplona; Sarkar, Sudeep

doi:10.1109/icb45273.2019.8987337

Cited by 25 publications

(18 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A second round of the UERC competition was held in 2019 [46]. The evaluation was performed using the experimental protocol and dataset partitions for training and testing as in [32].…”

Section: Related Workmentioning

confidence: 99%

“…Fine-tuning deep CNNs initially trained on the ImageNet dataset has become the de facto standard for building robust and high performing recognition models for vision and related domains [74], including ear recognition [43], [45], [46]. In order to apply fine-tuning we follow a two-step procedure.…”

Section: B Fine-tuningmentioning

confidence: 99%

“…In fact, state-of-the-art results in various recognition competitions and particularly the ImageNet challenge [50] are achieved through ensembles of deep networks. Furthermore, the best ear recognition performance in several conducted studies are achieved through ensembles of deep models [43], [75], [46].…”

Section: Ensemble Learningmentioning

confidence: 99%

See 2 more Smart Citations

Deep Convolutional Neural Networks for Unconstrained Ear Recognition

et al. 2020

View full text Add to dashboard Cite

This paper employs state-of-the-art Deep Convolutional Neural Networks (CNNs), namely AlexNet, VGGNet, Inception, ResNet and ResNeXt in a first experimental study of ear recognition on the unconstrained EarVN1.0 dataset. As the dataset size is still insufficient to train deep CNNs from scratch, we utilize transfer learning and propose different domain adaptation strategies. The experiments show that our networks, which are fine-tuned using custom-sized inputs determined specifically for each CNN architecture, obtain state-of-the-art recognition performance where a single ResNeXt101 model achieves a rank-1 recognition accuracy of 93.45%. Moreover, we achieve the best rank-1 recognition accuracy of 95.85% using an ensemble of fine-tuned ResNeXt101 models. In order to explain the performance differences between models and make our results more interpretable, we employ the t-SNE algorithm to explore and visualize the learned features. Feature visualizations show well-separated clusters representing ear images of the different subjects. This indicates that discriminative and ear-specific features are learned when applying our proposed learning strategies.

show abstract

“…A second round of the UERC competition was held in 2019 [46]. The evaluation was performed using the experimental protocol and dataset partitions for training and testing as in [32].…”

Section: Related Workmentioning

confidence: 99%

Section: B Fine-tuningmentioning

confidence: 99%

See 1 more Smart Citation

Deep Convolutional Neural Networks for Unconstrained Ear Recognition

et al. 2020

View full text Add to dashboard Cite

show abstract

“…They presented a comparative analysis of different covariates, such as occlusion, rotation, spatial resolution, and gallery size, on the recognition performance, and the obtained results highlighted insightful findings summarized in [43]. The second round of the challenge, UERC 2019 [44], evaluated 13 ear recognition techniques. The majority of the submitted approaches utilized deep learning techniques.…”

Section: Introductionmentioning

confidence: 99%

“…More interesting, around 50% of the approaches were based on multiple ear image representations due to limitations of a single model representation to capture the wide and complex appearance variability. Even though the obtained results show some improvements in the recognition rate when using ensembles or multiple descriptor combinations, the problem of recognizing humans from ear images under unconstrained conditions has not been solved and more research is needed [44]. The winner of the challenge was ScoreNet-5 [45] which utilizes a fusion learning approach.…”

Section: Introductionmentioning

confidence: 99%

Ensembles of Deep Learning Models and Transfer Learning for Ear Recognition

Alshazly

Linse

Barth

et al. 2019

Sensors

View full text Add to dashboard Cite

The recognition performance of visual recognition systems is highly dependent on extracting and representing the discriminative characteristics of image data. Convolutional neural networks (CNNs) have shown unprecedented success in a variety of visual recognition tasks due to their capability to provide in-depth representations exploiting visual image features of appearance, color, and texture. This paper presents a novel system for ear recognition based on ensembles of deep CNN-based models and more specifically the Visual Geometry Group (VGG)-like network architectures for extracting discriminative deep features from ear images. We began by training different networks of increasing depth on ear images with random weight initialization. Then, we examined pretrained models as feature extractors as well as fine-tuning them on ear images. After that, we built ensembles of the best models to further improve the recognition performance. We evaluated the proposed ensembles through identification experiments using ear images acquired under controlled and uncontrolled conditions from mathematical analysis of images (AMI), AMI cropped (AMIC) (introduced here), and West Pomeranian University of Technology (WPUT) ear datasets. The experimental results indicate that our ensembles of models yield the best performance with significant improvements over the recently published results. Moreover, we provide visual explanations of the learned features by highlighting the relevant image regions utilized by the models for making decisions or predictions.

show abstract