Segmentation and Normalization of Human Ears Using Cascaded Pose Regression

Pflug, Anika; Busch, Christoph

doi:10.1007/978-3-319-11599-3_16

Cited by 5 publications

(10 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We observe the highest performance increase with the ResNet-50 model, where the Rank-1 score is improved by 50% due to the alignment and the Rank-5 score is increased by 27%. If we focus only on the Rank-1 recognition rates, we can see relative improvements of 41% for ResNet-18, 23% ResNet-101, 24% for ResNet-152, 14% for MobileNet (0.25), 31% for MobileNet (0.5) and 17% for MobileNet (1). Similar improvements can also be observed for the other two performance indicators.…”

Section: Ear Recognition Resultssupporting

confidence: 69%

“…To address this issue, we studied the ear-landmark detection and alignment tasks in this work and developed a framework capable of locating a large number of ear landmarks in input images with diverse characteristics and aligning the ears with a pre-defined shape template. We formulated ear-landmark F I G U R E 1 5 Average ears computed from the aligned ear images of the AWEx dataset using CPR-based [1] and SIFT + RANSAC [7] alignment. Compared to the average of the unaligned ears (third image in Figure 14), the average CPR (left) and SIFT + RANSAC (right) aligned ear are less crisp.…”

Section: Discussionmentioning

confidence: 99%

“…The first is based on the Cascaded Pose Regression (CPR) framework from Refs. [1,23] and the second is a combination of the SIFT keypoint detector and RANSACbased alignment, originally presented in Ref. [7].…”

Section: Geometric Normalisationmentioning

confidence: 99%

“…Ear recognition techniques have seen considerable improvements over the years and consequently contributed towards increased interest in automated ear recognition systems [1,2]. While many powerful recognition approaches, mostly based on deep learning, have been proposed in the literature recently, the problem of ear alignment has received comparably less attention, as also emphasised in recent surveys in this field.…”

Section: Introductionmentioning

confidence: 99%

“…Normalisation results on the AWEx dataset. For each row, the first (far left) image shows the unaligned ear cropped from the input images after ear detection, the second one shows the aligned CPRbased[1] version, the third shows the image aligned with SIFT and RANSAC[7] and the fourth (far right) shows the proposed alignment.F I G U R E 1 4Average ears computed from unaligned and aligned ear images of the ITWE (left) and AWE (right) datasets. Note that the proposed geometric normalisation contributes towards better alignment of specific ear structures, which is reflected in the more crisp average appearance.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Efficient ear alignment using a two‐stack hourglass network

2023

View full text Add to dashboard Cite

Ear images have been shown to be a reliable modality for biometric recognition with desirable characteristics, such as high universality, distinctiveness, measurability and permanence. While a considerable amount of research has been directed towards ear recognition techniques, the problem of ear alignment is still under‐explored in the open literature. Nonetheless, accurate alignment of ear images, especially in unconstrained acquisition scenarios, where the ear appearance is expected to vary widely due to pose and view point variations, is critical for the performance of all downstream tasks, including ear recognition. Here, the authors address this problem and present a framework for ear alignment that relies on a two‐step procedure: (i) automatic landmark detection and (ii) fiducial point alignment. For the first (landmark detection) step, the authors implement and train a Two‐Stack Hourglass model (2‐SHGNet) capable of accurately predicting 55 landmarks on diverse ear images captured in uncontrolled conditions. For the second (alignment) step, the authors use the Random Sample Consensus (RANSAC) algorithm to align the estimated landmark/fiducial points with a pre‐defined ear shape (i.e. a collection of average ear landmark positions). The authors evaluate the proposed framework in comprehensive experiments on the AWEx and ITWE datasets and show that the 2‐SHGNet model leads to more accurate landmark predictions than competing state‐of‐the‐art models from the literature. Furthermore, the authors also demonstrate that the alignment step significantly improves recognition accuracy with ear images from unconstrained environments compared to unaligned imagery.

show abstract

Section: Ear Recognition Resultssupporting

confidence: 69%

Section: Discussionmentioning

confidence: 99%

Section: Geometric Normalisationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations