rmast

git bisect bad 65a077d3e9cd3be1f2ef97b84bed9f5649722775 is the first bad commit commit 65a077d3e9cd3be1f2ef97b84bed9f5649722775 Author: Robert Sachunsky Date: Mon Aug 24 19:01:03 2020 +0200 FindAndRemoveLines/FindVerticalAlignment: decrease fixed vline min length When detecting vertical...

This commit is either introducing the error, or shifting it into sight.

Seen the same, with a fresh built torch 1.9.1. I'll continue using newer torch-versions.

I forked for these fixes. I didn't include IMAP-functionality to keep the testset in file-format. Apart from the concatenation of broken lines, there also is a 'recurse'-function that doesn't recurse...

> Interesting. > > Is it consistent or once in a while? I repeated it, with same stall after the first epoch. By the way, the console is still responding,...

Yes, unfortunately DjVu-development of the segmenter has been stalled during the attempts to make it pay off by software patents which by now have expired. Don't know if those old...

I watched your demo-scan. It has the same wrong inversions as I have. The name of your pullrequest doesn't make sense.

I watched your PR. The original routine, which was not fully python3 did this conversion from this source picture: https://user-images.githubusercontent.com/3341558/175789293-f39ddfdb-6f3e-4598-8d16-80a1f4a88b36.jpg  Your version gives this picture:  You can...

When I run the original matlab code in Octave I get a better result, however missing most of the dots of the characters, so this jasonlfunk-translation does not fully represent...

When testing on [this nearly perfect image](https://user-images.githubusercontent.com/3341558/175789293-f39ddfdb-6f3e-4598-8d16-80a1f4a88b36.jpg) I still see a lot of things to improve to approach that quality. The jbig2-picture needs improvement. Not only the images should contain...