@DhruvJ

First Devlog

My Project idea seems really basic and simple and I thought so to!

so, I broke down the idea into 3 parts:
1)Pre Processing (turning to grayscale and popping out the notes more)
2)actual CV model
3)converting and displaying into text

So i began with the first part and realised there are so many factors

1)Resolution: -Photo needs resolution to make out actual keystrokes
2)Aspect ratio and resizing (I wasn’t aware, but models can only take in a fixed size)
3)Inconsistent Skew and orientation could really affect recognition of the same stroke
4)Different lightings in photos: I realized preprocessing on one image can be perfect but be blurry on another
5) fine turning the model

So, in this devlog I was learning about different factors do account for and realised that even jpeg and png can have different outputs

my first version works really bad in my attempt to making the notes pop

I converted it to grayscale which wasn’t enough so I tried a per-pixel intensity remap CLAHE which I still don’t understand how it works but the other methods with gaussian blur and median type blur wasn’t helpful.

Tasks accomplished:

1. Made a schematic.
2. Made an enclosure for the Radio.

The radio is supposed to powered by the phone and use the phones microphone and speaker as the interface, the goal is to practically compare ESP-NOW communication protocol and the 2.4ghz transceiver module and adaptively transition or switch to either protocol but there is a caveat, they communicate at the same frequency band and can interfere but we can only test it to know more