NLP-OCR

Small tasks carried on topics of NLP and OCR using deep learning. The purpose of the project was to reconstruct fazed old documents using deep learning.

-- OCR

• Using CV documentation methods like shape detecting using contours and changing perspective to convert image such as it is easy to scan the letters.
• For the OCR part document_basic.png is the original file and edge_basic.png is the image with edges only, opening.png is the edges image after an attempt for better boundaries, warped.png image is the changed perscpetive image after getting the boundaries from opening.png and applying them on original image. final.png is the final image which we got after finetuning the warped image.

-- NLP

• Using a file containg thousands of reviews and their sentiment as a database I created elementary elementary neural network by logistic regression build a simple classifier model on Text to classify sentences based on their sentiment, using fundamental "bag of words" method.
• CVI_NLP - sheet1.pdf contains various success rates on varying certain parameters.