Peer-reviewedCC BYJournal of imaging6 Oct 2026

Assessment of Convolutional Neural Network and Vision Transformer Architectures for Structural Surface Classification.

Bonginkosi A Thango¹
Sipho G Thango^2,3

¹Department of Electrical and Electronic Engineering Technology, University of Johannesburg, Johannesburg 2092, South Africa.
²Discipline of Civil Engineering, University of KwaZulu Natal, Durban 4041, South Africa.
³National Home Builders Registration Council, Durban 4001, South Africa.

Abstract

We evaluate ten convolutional neural networks (CNNs) and ten transformer/attention architectures for nine-category surface classification using 41,756 balanced StructDamage images. A frozen ImageNet-pretrained ResNet18 encoder and learned head use 33,352 training, 4202 validation and 4202 test images grouped by inferred augmentation family. Test accuracy is 98.88%, macro-F1 is 0.9881, and the 95% component-bootstrap accuracy interval is 98.42-99.26%. Content screening retains 4102 test images and 98.85% accuracy. Nevertheless, cross-partition exact matches and candidate source-prefix overlap prevent treating these results as independent-site validation. The separate random-initialization screen uses 1800 training images, 900 validation images, the common test set, 64 × 64 inputs and five epochs. DenseNet121 leads with 88.34% accuracy; XCiT-Tiny leads the transformer/attention group with 84.75% accuracy. Eighteen models select the final epoch, so the comparison does not establish convergence-equivalent superiority. With 5400 training images, all twenty accuracies increase and the between-budget rank correlation is 0.850. Component-aware comparisons and wall/deck error analysis qualify the ranking. The findings concern surface-category recognition, not defect localization, damage severity or structural safety.

Comparative and explainable deep learning for cherry (Prunus avium L.) cultivar classification using convolutional neural networks (CNN) and transformer architectures
Bünyamin Demir, Fatih Ünal, Ömer Türk +5
Scientific Reports30 Oct 2026Peer-reviewedCC BY 4.0Crossref
MedFuse: dual-stream fusion of convolutional and vision transformer-based features for enhanced medical image classification
Yajing Ren, Ling Hai, Zheng Gu +1
Frontiers in medicine16 Oct 2026Peer-reviewedCC BYFull textGeneral medicineEurope PMC
Person-independent and cross-dataset evaluation of convolutional and recurrent architectures for eye movement classification
Ahmet Akkaya
Scientific Reports11 Oct 2026Peer-reviewedCC BY 4.0Full textCrossref
Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification
Viraj Wankhade, Vrushali Paratkar, Bhuvan Bhiogade +2
Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification25 Oct 2026PreprintCC BY 4.0Zenodo
Interpretable attention-deficit hyperactivity disorder diagnosis from functional magnetic resonance imaging via hybrid convolutional neural network–transformer learning and multi-level graph-based comprehensibility
R. Harini, O. Uma Maheswari, K. Lakshmi Priya
Scientific Reports5 Oct 2026Peer-reviewedCC BY-NC-ND 4.0Full textPsychiatry & mental healthCrossref

Cite

BibTeX RIS

BibTeX

@article{thango2026assessment,
  title = {{Assessment of Convolutional Neural Network and Vision Transformer Architectures for Structural Surface Classification.}},
  author = {Bonginkosi A Thango and Sipho G Thango},
  journal = {Journal of imaging},
  year = {2026},
  month = oct,
  doi = {10.3390/jimaging12100487},
  url = {https://europepmc.org/article/MED/42930638},
  note = {CC BY},
}

RIS

TY  - JOUR
TI  - Assessment of Convolutional Neural Network and Vision Transformer Architectures for Structural Surface Classification.
AU  - Bonginkosi A Thango
AU  - Sipho G Thango
PY  - 2026
DA  - 2026/10/06
JO  - Journal of imaging
DO  - 10.3390/jimaging12100487
UR  - https://europepmc.org/article/MED/42930638
AB  - We evaluate ten convolutional neural networks (CNNs) and ten transformer/attention architectures for nine-category surface classification using 41,756 balanced StructDamage images. A frozen ImageNet-pretrained ResNet18 encoder and learned head use 33,352 training, 4202 validation and 4202 test images grouped by inferred augmentation family. Test accuracy is 98.88%, macro-F1 is 0.9881, and the 95% component-bootstrap accuracy interval is 98.42-99.26%. Content screening retains 4102 test images and 98.85% accuracy. Nevertheless, cross-partition exact matches and candidate source-prefix overlap prevent treating these results as independent-site validation. The separate random-initialization screen uses 1800 training images, 900 validation images, the common test set, 64 × 64 inputs and five epochs. DenseNet121 leads with 88.34% accuracy; XCiT-Tiny leads the transformer/attention group with 84.75% accuracy. Eighteen models select the final epoch, so the comparison does not establish convergence-equivalent superiority. With 5400 training images, all twenty accuracies increase and the between-budget rank correlation is 0.850. Component-aware comparisons and wall/deck error analysis qualify the ranking. The findings concern surface-category recognition, not defect localization, damage severity or structural safety.
SN  - 2313-433X
ER  -

Abstract

Related

Cite