Adding geodesic information and stochastic patch-wise image prediction for small dataset learning

Hammoumi, Adam; Moreaud, Maxime; Ducottet, Christophe; Desroziers, Sylvain

doi:10.1016/j.neucom.2021.01.108

Cited by 6 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our case, 30 images of support A were manually segmented with respect to the two components (i.e., inclusions and matrix) to build the training data set. Since this number of samples is insufficient to train the model properly, a patch training strategy introduced by Hammoumi et al 34 was used. This approach consists of decomposing an image into several regions on which the training will be based.…”

Section: Methodsmentioning

confidence: 99%

“…At the inference time, the predicted segmented image is fully assembled by a stochastic process drawing the patches with random coordinates. This stochastic assembly of patches, introduced as a stratified sampling strategy by Hammoumi et al, 34 allows to avoid edge effects at the border of patches when they are regularly distributed. The U-Net architecture adapted to our data set is shown in Figure 3 .…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

SEM Image Processing Assisted by Deep Learning to Quantify Mesoporous γ-Alumina Spatial Heterogeneity and Its Predicted Impact on Mass Transfer

Głowska,

Jolimaitre,

Hammoumi

et al. 2024

J. Phys. Chem. C

Self Cite

View full text Add to dashboard Cite

The pore network architecture of porous heterogeneous catalyst supports has a significant effect on the kinetics of mass transfer occurring within them. Therefore, characterizing and understanding structure–transport relationships is essential to guide new designs of heterogeneous catalysts with higher activity and selectivity and superior resistance to deactivation. This study combines classical characterization via N2 adsorption and desorption and mercury porosimetry with advanced scanning electron microscopy (SEM) imaging and processing approaches to quantify the spatial heterogeneity of γ-alumina (γ-Al2O3), a catalyst support of great industrial relevance. Based on this, a model is proposed for the spatial organization of γ-Al2O3, containing alumina inclusions of different porosities with respect to the alumina matrix. Using original, advanced SEM image analysis techniques, including deep learning semantic segmentation and porosity measurement under gray-level calibration, the inclusion volume fraction and interphase porosity difference were identified and quantified as the key parameters that served as input for effective tortuosity factor predictions using effective medium theory (EMT)-based models. For the studied aluminas, spatial porosity heterogeneity impact on the effective tortuosity factor was found to be negligible, yet it was proven to become significant for an inclusion content of at least 30% and an interphase porosity difference of over 20%. The proposed methodology based on machine-learning-supported image analysis, in conjunction with other analytical techniques, is a general platform that should have a broader impact on porous materials characterization.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

SEM Image Processing Assisted by Deep Learning to Quantify Mesoporous γ-Alumina Spatial Heterogeneity and Its Predicted Impact on Mass Transfer

Głowska,

Jolimaitre,

Hammoumi

et al. 2024

J. Phys. Chem. C

Self Cite

View full text Add to dashboard Cite

show abstract

“…A binary morphological operation was performed to keep only connected components completely contained in the image and remove all those in contact with the border of the image, since in this case the full length of the microfibers is impossible to quantify. Then, a morphological analysis algorithm [40] was used to determine the geodesic length of each connected component in the resulting binary image, which renders the length of each fiber in pixel. This value can then be converted to metric units (pixel size of 2.73 µm/pixel).…”

Section: Counting Methodsmentioning

confidence: 99%

Quantification and Morphological Characterization of Microfibers Emitted from Textile Washing

Dreillard¹,

Barros²,

Rouchon³

et al. 2022

SSRN Journal

View full text Add to dashboard Cite

“…To address the issue of missing inter-slice information and bridge the gap further between 2D and 3D strategies, a distance transform was applied to the initial volume before slicing. This approach, introduced in [15], allows to enhance 2D images by adding geodesic information in non-textured areas. The ADS-net, trained on augmented 2D images by 3D geodesic information, can be considered somehow as a 2.5D network.…”

Section: Network Architecturementioning

confidence: 99%

Accelerating a Morphology-Preserving Adsorption Model by Deep Learning

Hammoumi

Moreaud

Jolimaître

et al. 2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Physico-chemical simulations are often based on time consuming statistical microstructures models. Recent works show that mathematical morphology and non-linear image processing operators could provide less time-consuming alternatives [1]. Even though these models allow a significant improvement of computation cost, the solving time for a single run on a large microstructure may still require many hours. This paper proposes a two-step approach to overcome this issue in the field of gas adsorption modeling. The first step is to encompass physico-chemical calculation data into 3D adsorption maps. The second step is to train an encoder-decoder convolutional neural network on these maps, to estimate adsorption maps. Using this new approach, the computation time of the adsorption curve for a single run (512 voxels) was reduced by a factor of more than 30.

show abstract

Adding geodesic information and stochastic patch-wise image prediction for small dataset learning

Cited by 6 publications

References 18 publications

SEM Image Processing Assisted by Deep Learning to Quantify Mesoporous γ-Alumina Spatial Heterogeneity and Its Predicted Impact on Mass Transfer

SEM Image Processing Assisted by Deep Learning to Quantify Mesoporous γ-Alumina Spatial Heterogeneity and Its Predicted Impact on Mass Transfer

Quantification and Morphological Characterization of Microfibers Emitted from Textile Washing

Accelerating a Morphology-Preserving Adsorption Model by Deep Learning

Contact Info

Product

Resources

About