MobileRead Forums - View Single Post - Selective binarization and re-encoding program for use with KoReader

LegeApps · Yesterday, 05:07 AM

Update--

Lege is now opensource again. Version is 1.4.76. OCR thorough has been improved, along with renderer optimizations.

Additions to the repo since last time it was public -

Lege-viewer - A PDF reader in development. has pagedown/pageup that snaps to an exact line rather than going by page dimensions, which is something I always wanted in a viewer. Could benefit from lots of improvement. pull requests welcomed..

Lege-pdf - the aforementioned renderer. Can see that there's no secret sauce, it's just a better architecture. "DocumentSnapshot". Can benefit from nearly endless optimizations. Pdfium still gets weekly updates after 10 years, same thing here.

Lege-document-ocr - this is a custom designed PDF OCR pipeline, was intended for a business application. Can OCR 40 pages a second using TensorRT paddle or windows OCR + the renderer. Needs more improvement, spellcheck etc.

Lege-android - android version, fully functional.

The reason for opensourcing again is just that it should be, and selling any particular component also requires supporting it and a lot of stuff I just don't want to do. All Lege originating code is AGPL3 which means you can use it in any project, you just have to display the license, and your project has to be opensource. The A in the AGPL means there is no workaround for using it on a webserver without having to opensource it. It's all really a good faith thing, who's going to police it..that was one of my hangups, but I got over it.

Any collab is welcomed, or just keep using the program like you were. Back to my other projects..

Yesterday, 05:07 AM	#54
LegeApps Enthusiast Posts: 26 Karma: 10 Join Date: Jun 2026 Device: Kobo Clara BW	Update-- Lege is now opensource again. Version is 1.4.76. OCR thorough has been improved, along with renderer optimizations. Additions to the repo since last time it was public - Lege-viewer - A PDF reader in development. has pagedown/pageup that snaps to an exact line rather than going by page dimensions, which is something I always wanted in a viewer. Could benefit from lots of improvement. pull requests welcomed.. Lege-pdf - the aforementioned renderer. Can see that there's no secret sauce, it's just a better architecture. "DocumentSnapshot". Can benefit from nearly endless optimizations. Pdfium still gets weekly updates after 10 years, same thing here. Lege-document-ocr - this is a custom designed PDF OCR pipeline, was intended for a business application. Can OCR 40 pages a second using TensorRT paddle or windows OCR + the renderer. Needs more improvement, spellcheck etc. Lege-android - android version, fully functional. The reason for opensourcing again is just that it should be, and selling any particular component also requires supporting it and a lot of stuff I just don't want to do. All Lege originating code is AGPL3 which means you can use it in any project, you just have to display the license, and your project has to be opensource. The A in the AGPL means there is no workaround for using it on a webserver without having to opensource it. It's all really a good faith thing, who's going to police it..that was one of my hangups, but I got over it. Any collab is welcomed, or just keep using the program like you were. Back to my other projects..