Automatic Ear Landmark Localization, Segmentation, and Pose Classification in Range Images

Lei, Jiajia; You, Xinge; Abdel-Mottaleb, Mohamed

doi:10.1109/tsmc.2015.2452892

Cited by 31 publications

(6 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The study segmented seven major areas inside the auricle including the helix, the antihelix, and the concha cavity. Similar work is presented by Lei et al [19] An ear tree-structured graph (ETG) was proposed and a 3-D flexible mixture model was trained to locate 18 landmarks of the auricle anatomy in 3D. However, no auriculotherapy-related landmarks are localized in both [18] and [19].…”

Section: Related Workmentioning

confidence: 96%

“…Similar work is presented by Lei et al [19] An ear tree-structured graph (ETG) was proposed and a 3-D flexible mixture model was trained to locate 18 landmarks of the auricle anatomy in 3D. However, no auriculotherapy-related landmarks are localized in both [18] and [19]. Wen et al [20] used the ASM algorithm and 25 auricular subzones were divided according to the WFAM standard.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning-Based Auricular Point Localization for Auriculotherapy

Sun

Dong

Li³

et al. 2022

IEEE Access

View full text Add to dashboard Cite

Auriculotherapy is one of the main forms of treatment in Traditional Chinese Medicine, whose potential as an alternative medicine for both health evaluation and disease treatment has been reported in many cases. However, its efficacy highly relies on the accurate localization of auricular points, which are not easy to be remembered due to their complexity. To explore an efficient way of locating auricular points, this study proposed a deep learning-based method of automatically locating auricular points from auricular images. A self-collected dataset named EID was created for TCM auriculotherapy research, with 91 auriculotherapy-related landmark points manually annotated according to the Chinese national standardization. A deep neural network structure was trained for landmark detection, and a direction normalization module was proposed to compensate for the detection error caused by the difference between the left and right ears. The trained model was validated on dataset EID. An average NME of 0.0514±0.0023 was achieved, which outperformed similar works. In addition, a certain auricular area corresponding to the digestive system was segmented based on the localized landmarks, and the results were tested in real-time video streaming. The proposed work for both auricular landmark and area identification can be widely used in auriculotherapy education and applications.

show abstract

Section: Related Workmentioning

confidence: 96%

Section: Related Workmentioning

confidence: 99%

Deep Learning-Based Auricular Point Localization for Auriculotherapy

Sun

Dong

Li³

et al. 2022

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Third, at the data analysis level, after validation of the acquired data by using the Kaiser-Meyer-Olkin (KMO) test and the Bartlett test, 43 we applied principal factor extraction to reduce the dimensionality of the measured anthropometric data by projecting them onto a low-dimensional subspace, where each axis (factor) is a linear combination of the initial variables (items). [44][45][46][47] To provide a convenient measurement for users, we only found one representative item from each principal factor, and then used these items for body classification. Finally, the K-means clustering method 48,49 was used to classify each body part, in which the number of classes or clusters was determined using analysis of variance (ANOVA).…”

Section: Overview Of the Proposed Modeling Proceduresmentioning

confidence: 99%

A new parametric 3D human body modeling approach by using key position labeling and body parts segmentation

Cheng

Zeng

Bruniaux

et al. 2022

Textile Research Journal

View full text Add to dashboard Cite

The current three-dimensional human models used in the apparel industry are mostly rigid and lack semantic information on body positions and body parts. Therefore, it is difficult for designers to make accurate, fast and effective designs from these models. This paper proposes a new parametric three-dimensional human body model based on key position labeling and optimized body parts segmentation. First, by using experts’ professional knowledge, we manually realize accurate human body data measurements as well as their interpretation and classification, and extract relevant human body features. After deep analysis, measured data irrelevant to body shape have been excluded by designers. Furthermore, the relation between body shapes and body features have been modeled. Second, based on this relational model, we label key positions on the corresponding three-dimensional body model obtained by scanning and segmenting the whole three-dimensional human body into semantically interpretable body parts. In this way, two databases have been created, enabling us to identify features of all segmented body parts, whose combination corresponds to the whole body shape. Third, for a specific consumer, his/her personalized three-dimensional human model can be obtained by taking a very few number of body measurements on himself/herself, making an appropriate combination of the identified body parts, and adjusting parameters of all involved body parts. By comparing the proposed labeled and segmented three-dimensional human model and the existing human models through a number of experiments, the proposed model leads to more relevant results with high accuracy and high visual quality related to real human body shapes.

show abstract

“…5 Accuracy test for photogrammetry and depth camera acquisition 2. The spectrum of solutions that can be adopted and designed to approach this goal includes the use of threedimensional anatomical data, as well as 2D or hybrid solutions [15][16][17]. In addition, segmentation strategies, already used in the literature for different anatomical districts, can be investigated [18].…”

Section: Step 1a: Healthy Ear Acquisitionmentioning

confidence: 99%

A rapid prototyping approach for custom training of autologous ear reconstruction

Mussi

Servi

Facchini

et al. 2021

Int J Interact Des Manuf

View full text Add to dashboard Cite

Autologous ear reconstruction is the preferred treatment in case of partial or total absence of the patient external ear. This kind of surgery can be really challenging since precise replication of complex three-dimensional structure of the ear is crucial to provide the patients with aesthetically consistent reconstructed anatomy. Therefore, the results strongly depends on the “artistic skills” of the surgeon who is in charge to carry out a three-dimensional sculpture, which resembles the shape of a normal ear. In this context, the definition of a preoperative planning and simulation process based on the patient's specific anatomy may help the surgeon in speeding up the ear reconstruction process and, at the same time, to obtain better results, thus allowing a superior surgical outcome. In the present work the main required features for performing an effective simulation of the ear reconstruction are identified and a strategy for their interactive design and customization is devised with the perspective of a semi-automatization of the procedure. In detail, the paper provides a framework which start from the acquisition of 3D data from both a healthy ear of the patient (or, if not available e.g. due to bilateral microtia of the ear of one of his parents or from a template) and of costal cartilage. Acquired 3D data are properly processed to define the anatomical elements of the ear and to find, using nesting-based algorithms, the costal cartilage portions to be used for carving the ear itself. Finally, 3D printing is used to create a mockup of the ear elements and a prototype of the ear to be reconstructed is created. Validated on a test case, the devised procedure demonstrate its effectiveness.

show abstract

Automatic Ear Landmark Localization, Segmentation, and Pose Classification in Range Images

Cited by 31 publications

References 36 publications

Deep Learning-Based Auricular Point Localization for Auriculotherapy

Deep Learning-Based Auricular Point Localization for Auriculotherapy

A new parametric 3D human body modeling approach by using key position labeling and body parts segmentation

A rapid prototyping approach for custom training of autologous ear reconstruction

Contact Info

Product

Resources

About