Greetings from Kansas City, Missouri

A place to introduce yourself, and to meet other awesome people.

Moderator: peterZ

Post Reply
bpetty
Posts: 2
Joined: 07 Jun 2019, 21:21
Number of books owned: 0
Country: United States

Greetings from Kansas City, Missouri

Post by bpetty »

Hello everyone, my name is Brandon Petty.

I am interested in talking to more people about Document Image Processing so I thought this would be a good place to start.
I currently have been doing research into different Binarization techniques and have posted a github repo of my work:
https://github.com/brandonmpetty/Doxa
https://brandonmpetty.github.io/Doxa/WebAssembly

My goal is to wrap up that project in about another year, after implementing Howe's algorithm and building out my own CNN.
After that, I plan on experimenting with binary to vector graphics, document layout analysis, and eventually OCR of handwritten manuscripts.
I am not an academic, so I am learning as I go along.

I was drawn to this forum when I found out that someone here actually implemented Howe's algorithm. It sounds like there are a lot of creative minds here... capable of creating something tangible out of the otherwise abstract, so I am looking forward to talking to you all!
zbgns
Posts: 61
Joined: 22 Dec 2016, 06:07
E-book readers owned: Tolino, Kindle
Number of books owned: 600
Country: Poland

Re: Greetings from Kansas City, Missouri

Post by zbgns »

Hello Brandon,
Effects of binarization presented in form of two screenshots at your Github repository are really impressive. Do you have plans to add some interface to your algorithm for users that want to use it as standalone program?
bpetty
Posts: 2
Joined: 07 Jun 2019, 21:21
Number of books owned: 0
Country: United States

Re: Greetings from Kansas City, Missouri

Post by bpetty »

Hi zbgns!

I hope that output looks good... because that is the ground truth you are actually looking at :)
It has been a real struggle finding an algorithm that can capture the thin, faint, lines within some of the glyphs I am working with.
With the right parameters, ISauvola can get really close.

As for interfaces, right now I only have plans for two.
One is a command line demo app that lets you run all of the algorithms, with parameters, from the command line.
The other is the beta demo I have up for WebAssembly. I plan on updating it soon-ish so that you can supply your own source image and optional ground truth (gt is used only for calculating performance). Accessing this in-browser, in my opinion, seems like the best of both worlds. If you have any thoughts about it, I'd be interested. Experimenting with WebAssembly is new for me.

Thanks for reaching out. If I do anything interesting with the repo I'll post it on the forum. If there is anything you'd like me to look into, I'd be interested in trying it out. I might take a break from binarization for a while and try to convert some of these things into vector graphics... something kind of like Adobe ClearScan. But like I said, if I do anything creative, I'll try to post it on this forum. I am always looking for feedback and new ideas.
zbgns
Posts: 61
Joined: 22 Dec 2016, 06:07
E-book readers owned: Tolino, Kindle
Number of books owned: 600
Country: Poland

Re: Greetings from Kansas City, Missouri

Post by zbgns »

Hi Brandon,

Many thanks for sharing effects of your work with all. I'm convinced that not only me but many other people keep fingers crossed for implementation of something similar to Adobe ClearScan.

Cheers.
Post Reply