Table 15 Comparison of proposed work with literary works in Localization and Recognition accuracy
Works | Method | Dataset | Localization accuracy (%) | Recognition accuracy (%) |
|---|---|---|---|---|
Character recognition, documentation system to decode Inscription5 | 2D CNN + Tesseract | Tamil temple inscriptions | 70.00 | 77.7 |
Palm leaf inscription classification7 | Contour-based convex hull + MABN GLCM | Palm leaf inscription—Tamil | 87.5 | 85.3 |
Multilingual character segmentation8 | Projection profiles, graph distance + KNN | Proprietary Devanagari, CPAR dataset | 97.26 | 87.42 |
PNN and deep learning character recognition9 | ROI CCA + DCNN | Tulu palm leaf characters | 64.51 | 88.07 |
Handwritten character recognition10 | Vertical projection method + CNN | 1000 random words | 95.6 | 85.74 |
Structural representation-based handwritten character recognition11 | Z ordering method + SVM | HP India 2013 laboratory dataset | 88.30 | 90.30 |
Bounding box method-based accurate vehicle number detection12 | Sobel Edge Detector, median filtering, dilation, erosion, flood fill, bound boxes | 25 License plates in Indian style (black letters on white and yellow backgrounds) | 90.6 | 93.00 |
Kannada handwritten scripts16 | LBP features + KNN | Kannada Vijayanagara, Wadiyar, Hoysala handwritten scripts | 90.91 | 93.9 |
Optical character segmentation for touching characters17 | Horver method + DLA | Tamil palm leaf manuscripts | 90.63 | 92.43 |
Character-level street view text spotting22 | MSRD + ResNet50 | ICDAR 2013, ICDAR 2015, MSRA TD 500, SCUT, RCTW, ReCTS, Shopsign | 92.98 | 92.53 |
Telugu printed handwritten documents using iterative split analysis26 | Iterative split + Projections | Handwritten Telugu (2400 chars) | 95.37 | 91.17 |
Stroke extraction for offline handwritten mathematical expression28 | Noise reduction, stroke direction, stroke order normalization, stroke tracing, double-traced stroke restoration, Over segmentation technique | CROHME 2016 dataset | 85.00 | 65.65 |
Signary localization and recognition (Proposed) | DPB-Dynamic Profiling Bound Approach + SignaryNet | 11th century Brihadisvara Tamil inscriptions | 98.25 | 96.81 |
