звездочёт

Hi all. [Butteraugli](https://github.com/google/butteraugli) is too long. Use the [guetzli](https://github.com/google/guetzli) encoder (it has this metric).

Hi all. You can also use [rdopt](https://github.com/ImageProcessing-ElectronicPublications/rdopt) for similar purposes. [Discussion](https://github.com/mozilla/mozjpeg/issues/182)

❓ Maybe use `jpeg-recompress` based on the standard `libjpeg`: https://github.com/ImageProcessing-ElectronicPublications/jpeg-recompress ?

SDR33 may contain different values. Need rules for test files.

1 station, 1 landmark, and many polar points (SZB). Got it.

Formats: {"Geodimeter JOB" {.job}} {"Geodimeter ARE" {.are}} {"Sokkia set 4" {.scr}} {"Sokkia sdr" {.sdr}} {"Leica GSI" {.gsi}} {"TopCon GTS-700" {.700}} {"TopCon GTS-210" {.210}} {"Trimble M5" {.m5}} {"Nikon DTM-300" {.nik}} {"Geodat...

Update repo by @zsiki : https://github.com/zsiki/GeoEasy/commit/ae04dcc0133fb94e8f309324e606897271b1b146 * foif (*.mes) * geodimeter (*.are+*.job) * geoprofi (*.mjk) * geozseni (*.gjk) * leica (*.gsi+*.idx) * nikon (*nik,*.raw) * ruide (*.sdr) * sokkia (*.scr)...

I’m not sure that the neural network makes sense (it will give a result on what it was trained on). There are more standard (but no less expensive) solutions: https://github.com/zvezdochiot/inpaint-cimg

Not so simple. The main code (https://github.com/ZQPei/patchmatch_inpainting/blob/master/src/inpaint.cpp) is borrowed (link bottom README). I fixed `main.cpp` and around it, without getting into the main code. Perhaps @ZQPei climbed deeper.

Hi @zsiki . SVD is optional. A completely sane solution is obtained with the simplest normalization of variables (`Xn = X - M(X)`). For a detailed solution with Gaussian normalization...