NiveditJain / Simultaneous-Optimisation-of-Image-Quality-Improvement-and-Text-Content-Extraction-from-scanned-docu

B.Tech Project

Simultaneous-Optimisation-of-Image-Quality-Improvement-and-Text-Content-Extraction-from-scanned-documents

• Optimization of image quality using SRCNN and modifying it to get improved results for text content extraction.
• Modification of loss function in accordance with
• Developed a novel metrics for determining the OCR accuracy.

Experiments

• Develped a 2 layer SRCNN which gives better results in comparison with SRCNN with regularization on the cost of training time.
• Trained an SRCNN model over the results obtained from SRCNN with regularization.
• Trained SRCNN over OTSU of images.

Dataset Links

• Results with models
• Metric Results
• Complete Dataset
• Presentation
• Report

References

• G. Youm, S. Bae and M. Kim, "Image super-resolution based on convolution neural networks using multi-channel input," 2016 IEEE 12th Image, Video, and Multidimensional Signal Processing Workshop (IVMSP), Bordeaux, 2016, pp. 1-5, doi: 10.1109/IVMSPW.2016.7528224.
• S. Mujumdar, N. Gupta, A. Jain and D. Burdick, "Simultaneous Optimisation of Image Quality Improvement and Text Content Extraction from Scanned Documents," 2019 International Conference on Document Analysis and Recognition (ICDAR), Sydney, Australia, 2019, pp. 1169-1174, doi: 10.1109/ICDAR.2019.00189.

System Requirements

Python librarires -

• numpy
• openCV 4.4.0
• OS
• pytesseract
• tensorflow
• keras

---

Run following commands on Linux Terminal to run the application

git clone https://github.com/LezendarySandwich/Simultaneous-Optimisation-of-Image-Quality-Improvement-and-Text-Content-Extraction-from-scanned-docu.git

cd Simultaneous-Optimisation-of-Image-Quality-Improvement-and-Text-Content-Extraction-from-scanned-docu

cd 'Required model'

python3 run_model.py

---

Team
Nivedit Jain (B18CSE039)
Sanskar Mani (B18CSE048)

---