Instance segmentation of 3D woven fabric from tomography images by Deep Learning and morphological pseudo-labeling

Blusseau, Samy; Wielhorski, Yanneck; Haddad, Zyad; Velasco-Forero, Santiago

doi:10.1016/j.compositesb.2022.110333

Cited by 14 publications

(4 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Indeed, many works devoted to extracting the mesoscale textile model, employ image processing techniques and perform quantitative analyses therein [3,4,5]. The techniques range from "classical" methods such as clustering operations [6] or texture analysis [7], up to the more recent Deep Learning approaches [8,9,10,11,12]. Other approaches based on deformable models have also been explored [13,14].…”

Section: Introductionmentioning

confidence: 99%

Alignment of 3D woven textile composites towards their ideal configurations

Rubino,

Mendoza,

Wielhorski

et al. 2024

Computer Methods in Applied Mechanics and Engineering

Self Cite

View full text Add to dashboard Cite

Section: Introductionmentioning

confidence: 99%

Alignment of 3D woven textile composites towards their ideal configurations

Rubino,

Mendoza,

Wielhorski

et al. 2024

Computer Methods in Applied Mechanics and Engineering

Self Cite

View full text Add to dashboard Cite

“…There is often a lack of available public datasets in the textile field. In order to reduce expensive manual annotation, training samples can be obtained by generating pseudo-images [ 32 ] or pseudo-labels [ 33 ] when different slices are said to have similar cross-sectional structures. Zheng et al [ 34 ] proposed a new data augmentation algorithm to generate realistic artificial datasets.…”

Section: Introductionmentioning

confidence: 99%

Extraction of the Microstructure of Wool Fabrics Based on Structure Tensor

Zhu,

Ma,

Ding

et al. 2023

Sensors

View full text Add to dashboard Cite

The trends of “fashionalization”, “personalization” and “customization” of wool fabrics have prompted the textile industry to change the original processing design based on the experience of engineers and trial production. In order to adapt to the promotion of intelligent production, the microstructure of wool fabrics is introduced into the finishing process. This article presents an automated method to extract the microstructure from the micro-CT data of woven wool fabrics. Firstly, image processing was performed on the 3D micro-CT images of the fabric. The raw grayscale data were converted into eigenvectors of the structure tensor to segment the individual yarns. These data were then used to calculate the three parameters of diameter, spacing and the path of the center points of the yarn for the microstructure. The experimental results showed that the proposed method was quite accurate and robust on woven single-ply tweed fabrics.

show abstract

“…If there is no fibrous pattern in the tows, the Gray Level Co-Occurence Matrix can be used [21]. Recent works have shown the benefits of machine learning algorithms with this process [26][27][28]. They can be applied to CT-scans with lower resolutions, because they consider larger features of the reinforcement, such as the yarns shapes and boundaries which are difficult to take into account with previously cited methods.…”

Section: Introductionmentioning

confidence: 99%