A benchmark shows lightweight CNNs with transfer learning reach 92-99% accuracy on Arabic handwritten character datasets, but the reported best-model rankings are undermined by best-fold selection and internal contradictions.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Leveraging Transfer Learning and Mobile-enabled Convolutional Neural Networks for Improved Arabic Handwritten Character Recognition
A benchmark shows lightweight CNNs with transfer learning reach 92-99% accuracy on Arabic handwritten character datasets, but the reported best-model rankings are undermined by best-fold selection and internal contradictions.