• Publications
  • Influence
Linear Curve Fitting-Based Headline Estimation in Handwritten Words for Indian Scripts
TLDR
The proposed method is able to detect headlines in skewed word images and provides accurate result even when the headline is discontinuous or mostly absent, and is compared with a recent work to show the efficacy of the proposed methodology.
Segmentation-based recognition system for handwritten Bangla and Devanagari words using conventional classification and transfer learning
TLDR
The authors propose a system that first detects and corrects skew present in Bangla and Devanagari handwritten words, estimates the headline, and further segments the words into meaningful pseudo-characters to provide the final result.
A Study on the Effect of CNN-Based Transfer Learning on Handwritten Indic and Mixed Numeral Recognition
TLDR
This study uses readily available pre-trained Convolutional Neural Network architectures on four different Indic scripts, viz.
Linear Regression-Based Skew Correction of Handwritten Words in Indian Languages
TLDR
This work proposes a method that uses linear curve fitting for estimating and correcting skew present in handwritten words and efficiently detects and corrects skew in four Indian languages, namely Bangla, Hindi, Marathi and Panjabi.
Finding the optimum classifier: Classification of segmentable components in offline handwritten Devanagari words
TLDR
An approach that finds the number of components present in a word image by tuning a range of classifiers to see if the component requires further segmentation or not and providing an accuracy of 98.63%.
A novel approach towards segmentation of connected handwritten numerals
TLDR
The proposed method can effectively separate the isolated and connected numerals in any unconstrained handwritten numeral string and is tested on NIST Database 19 dataset and observed to provide an accuracy of 95.84%.
A fuzzy and contour-based segmentation methodology for handwritten Hindi words in legal documents
TLDR
A character segmentation method that identifies the different zones of a word image and utilizes a fuzzy function for estimating the headline pixels and further uses the outer contour of the word along with the estimated headline pixels to segment the upper and lower modifiers, and meaningful constituent characters is proposed.
What the user does not want?: query reformulation through term inclusion-exclusion
TLDR
Experiments conducted on standard datasets like TREC, ROBUST, WT10G demonstrate that the proposed techniques yield substantial performance gain, often being statistically significant.
...
1
2
...