NAME: RAVULA SHIVAKUMAR GMAIL: ravula.shivakumar11@gmail.
com
NOTES OF WEEK-1
Day 1:
BFSI
Banking Financial services and insurances
Banking - transactional data
Financial services- Investments , stocks
Insurance - risk , it provides financial help.
Ocr-optical character recognition
Can be used to extract data from images or any document
Problems
Document management (digitalize documentation)
Cybersecurity threats
Fraud detection
RBI is an authority which regulate rules and supervise banks
It imposes huge fines ,if banks do not follow the regulations
Day 2 :
Ocr - convert into machine readable text
Flow of ocr
Take the image
Preprocess the image
Remove noice , segmentation-word ,character,sentence segmentation
Pattern finding
Post preprocessing-cross validation or verification of pixels
Image pre processing
Chracter recognition
Post preprocessing
Day 3:
Feature Lossless Compression Lossy Compression
Definition Reduces file size without losing any Reduces file size by discarding
data some data
Data Recovery Original data can be perfectly restored Original data cannot be fully
recovered
Quality No loss in quality Loss of quality due to data
removal
Compression Lower (less size reduction) Higher (more size reduction)
Ratio
Usage Text, medical images, software files Images, audio, video
Examples PNG JPEG
JPEG (Joint Photographic Experts Group) is a widely used image format that uses lossy
compression to reduce file size while maintaining good visual quality.
Png maintains a high definition. It's better to use png in ocr.
PNG (Portable Network Graphics) is a popular lossless image format known for high-quality
images and transparency support.
Key Features of PNG:
Lossless Compression – Retains all image data without quality loss.
Supports Transparency – Can have transparent or semi-transparent backgrounds.
Higher File Size – Larger than JPEG due to no data loss.
Best for Graphics & Web – Used for logos, web design, and images needing transparency.
A GIF (Graphics Interchange Format) is an image format that supports animation and lossless
compression.
Key Features of GIFs:
Supports Animation – Can store multiple frames to create short looping animations.
Lossless Compression – Maintains image quality but has a 256-color limit.
Transparency Support – Allows one color to be transparent.
Widely Used – Common for memes, stickers, and short clips.
Pdf - if only text present in it high accuracy of ocr
If both text and images are present accuracy of ocr degrades
Task 1 : Download and gather documents of bfsi sector
Ocr machines will perform task(tesseract-ocr)
Day 4 :
Preprocessing
Before performing OCR, preprocessing enhances accuracy.
Grayscale Conversion – Convert the image to grayscale to reduce noise.
Binarization – Convert the image to black and white example Otsu’s Thresholding.
Noise Reduction – Apply filters like Gaussian Blur or Median Blur.
Morphological Operations – to enhance text.
Deskewing – Corrects skewed text.
Contrast Enhancement – Adjust brightness and contrast.
Tasks
Perform all the preprocessing tasks
