Anonymous ID: fc90ed Sept. 18, 2018, 3 p.m. No.3077505   ๐Ÿ—„๏ธ.is ๐Ÿ”—kun   >>7640

>>3077427

Some more diverse examples hoping to foil Rosetta.

A couple of these I just lifted the text to a new layer on an existing meme, then used gimp filters to scramble/distort the edges.

I woke up this morning thinking make the text camo and made those new today.

Anonymous ID: fc90ed Sept. 18, 2018, 3:03 p.m. No.3077567   ๐Ÿ—„๏ธ.is ๐Ÿ”—kun

>>3077476

The software scans the image looking for rows of pixels that are aligned. They will look for contrast and edges. Once it determines an approximate baseline or orientation, they will start doing pattern matches on groups of pixels against known letter forms.

Did you know the Post Office has been using OCR to read handwritten addresses for, like, 2-3 decades now? Some get rejected and read by humans but the Optical Character Recognition software is, at this point, quite sophisticated.

I had a program that could read scans of old sheet music (notes on a staff) and "play" the musical notes. It wasn't perfect but the freeware version I had was pretty damn impressive and needed only minor corrections if the sheet music scans didn't have too many faults.

Anonymous ID: fc90ed Sept. 18, 2018, 3:08 p.m. No.3077640   ๐Ÿ—„๏ธ.is ๐Ÿ”—kun   >>8040

>>3077505

>>3077427

One more set of examples. I was playing around with visual contrast and confusing backgrounds using gimp layers and various blend filters to produce the text effects you see here.

Again, we don't know if these approaches are going to work. This is a buckshot approach of trying everything we can think of, and we will need some feedback after these memes encounter Rosetta.

Anonymous ID: fc90ed Sept. 18, 2018, 3:11 p.m. No.3077699   ๐Ÿ—„๏ธ.is ๐Ÿ”—kun

>>3077627

I believe they will be able to read this text.

I believe the word #fisa will trigger a ban.

But I could be wrong. We need to experiment with everything we can think of, and some of our efforts will work. We just don't know which ones yet.

Anonymous ID: fc90ed Sept. 18, 2018, 3:37 p.m. No.3078108   ๐Ÿ—„๏ธ.is ๐Ÿ”—kun

>>3078033

The Post Office's OCR reads every envelope, handwritten or not. Has been reading them for 20-30 years. Some get rejected and read by a human but not many.

If the handwriting is on a confused background then you'd have a greater chance of confusing Rosetta.

Play around and try stuff. That's all we can suggest. We need feedback as memes are deployed to discover what works against a SPECIFIC algorithm that has OCR capabilities and SPECIFIC logic programmed.