Neural Network OCR Bounding Box Merging for Handwritten Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Optical Character Recognition (OCR) techniques face difficulties in identifying words and sentences, especially in script-like languages or handwritten text, as they struggle to accurately translate graphical information into editable data.
Innovation Solution
A system that uses machine learning networks to determine image blobs, group them, and identify words by generating bounding boxes, which are then fed into neural networks for transcription, employing heuristics to optimize the identification and merging process, allowing for the recognition of words in various scripts and fonts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional OCR techniques are used to recognize text in input files, then text translation to editable data is achieved, but identification of words and sentences especially in script-like languages or handwritten text is difficult or unable
Solution Approach 1:
The system segments the text recognition process into multiple specialized components: an image processing module that extracts image blobs and generates bounding boxes, a neural network module that performs character classification, and a post-processing module that assembles recognized characters into words and sentences. This segmentation allows each module to be optimized for its specific function, improving overall accuracy for diverse text formats including handwritten and script-like languages
Solution Approach 2:
The system employs parameter changes by using a trained neural network with adjustable weights and biases that are optimized through training on diverse datasets including various scripts and handwritten texts. The neural network dynamically adjusts its parameters to adapt to different text formats, enabling reliable recognition across multiple languages and writing styles while maintaining high accuracy
2Measurement precision
If machine learning networks are used to identify words in images, then transcription accuracy is improved, but system complexity increases
Solution Approach 1:
The system introduces an intermediary image processing module that prepares input data for the neural network by extracting image blobs and generating bounding boxes. This intermediary layer simplifies the neural network's task by providing pre-processed, structured input, thereby maintaining high word identification accuracy while reducing the overall system complexity through modular design and clear separation of concerns
3Reliability
If bounding boxes are generated and merged to identify image data portions portraying words, then word identification capability is enhanced, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary actions by generating and merging bounding boxes before neural network inference to pre-segment the image into potential word regions. This pre-processing step improves word boundary detection accuracy by providing the neural network with focused input regions, while the efficient merging algorithm minimizes additional processing time by using optimized spatial operations to combine overlapping bounding boxes
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for neural network-based optical character recognition. An embodiment of the system may generate a set of bounding boxes based on reshaped image portions that correspond to image data of a source image. The system may merge any intersecting bounding boxes into a merged bounding box to generate a set of merged bounding boxes indicative of image data portions that likely portray one or more words. Each merged bounding box may be fed by the system into a neural network to identify one or more words of the source image represented in the respective merged bounding box. The one or more identified words may be displayed by the system according to a standardized font and a confidence score.