I have packaged this project. It enables to be easy to install and use as CLI commands (depix, genpixed). And I added type hints in mypy-strict mode.

If you don't mind, please publish this on PyPI!

ref: https://packaging.python.org/en/latest/tutorials/packaging-projects/#uploading-the-distribution-archives

This PR will add a Dockerfile and minor updates.

docker run -it --rm -v $(pwd):/data depix -p images/testimages/testimage3_pixels.png -s images/searchimages/debruinseq_notepad_Windows10_closeAndSpaced.png -o /data/output.png

INFO:root:Loading pixelated image from images/testimages/testimage3_pixels.png
INFO:root:Loading search image from images/searchimages/debruinseq_notepad_Windows10_closeAndSpaced.png
INFO:root:Finding color rectangles from pixelated space
INFO:root:Found 116 same color rectangles
INFO:root:86 rectangles left after moot filter
INFO:root:Found 1 different rectangle sizes
INFO:root:Finding matches in search image
INFO:root:Removing blocks with no matches
INFO:root:Splitting single matches and multiple matches
INFO:root:[10 straight matches | 76 multiple matches]
INFO:root:Trying geometrical matches on single-match squares
INFO:root:[15 straight matches | 71 multiple matches]
INFO:root:Trying another pass on geometrical matches
INFO:root:[17 straight matches | 69 multiple matches]
INFO:root:Writing single match results to output
INFO:root:Writing average results for multiple matches to output
INFO:root:Saving output image to: /data/output.png

When running depix with gamma correction on any image, you would get UnboundLocalError: local variable 'bb' referenced before assignment. This is caused by a typo when assigning colour component variables to pixels as the blue component (bb) is never assigned a value before it is used.

Add a section to describe common failure modes of the algorithm. Problem (1) is explored in genpixed.py, and it shouldn't be too hard to fix with an additional pixlating-mode switch. Problem (2) is much harder.

(Of course, this doesn't mean taking screenshots on a mac makes stuff safer.)

This project [unfortunately] has a pretty terrible README for newcomers.

At first, I didn't know how to use this. I hadn't a clue on what to put for the -s flag. Literal guesswork lead me to some form of a result. :(

What I've Done

• Fixed random spacing issues with paragraphs.
• Added a coherent installation and usage guide.
• Organised miscellaneous paragraphs which all seemed to fit nicely in an About section.
• Changed the link about De Bruijn sequences to one from Wikipedia, for a more credible and detailed description of what a De Bruijn sequence is.
• Upgraded understandability and overall quality of how to create your own search image, this helps for newcomers to understand how the magic of Depix works. ;)
• Removed the part on the so-called "dependency issue". It's nothing. You didn't tell people to run pip install -r requirements.txt for your project. Python can't run code without it's dependencies. I also added the fix to the Installation section of the README.

Thanks to open source, Depix will be better than ever!

Motivation

When reading the code, sometimes it's hard to know the data types or the meaning of variables. For example, what's the type of rectangleMatches? One would need to first go to depix.py and found that it was generated from findRectangleMatches(). After that, he or she knew it was of type dict, but what was the key and values?

Another motivation is that, when I tried out this tool the first time, the target image I used was inappropriate. Within that image, each 14x14 block was supposed to be of same color, but due to some lossy compression, each were instead composed of tiny little blocks as shown below:

I got stuck on this screen for over an hour:

Therefore, I think it would be better if the program could detect that there was too many variants on block size, which is what I added in depix.py line 47. The additional logging feature in line 109 and line 147 were also for addressing this problem by including more verbosity.

Feel free to edit if you think it's too verbose or redundant.

My name is Leo, a CS student at Brown building an automated platform for editable codebase documentation.

I generated documentation for this repository and placed a link in the README. Please feel free to try it out and let me know what you think!

Thank you for this awesome tool :thumbsup:

Description

To make the dependencies easier to install, i added a requirements.txt file.

What do you think about it?

There now exists an implementation of the HMM-based approach mentioned in the last section of the readme. I've added a reference to it.

+++Disclaimer: I'm the author of the implementation +++

Two lines of the README are indented, which makes GitHub think it's some code that is embedded. I removed the indent, which restores the text back to the dot points that they are. ;)

Description

• Criteria for checking neighbor blocks in findGeometricMatchesForSingleResults (#39).
• Possibly bugfix:
  • There are four loops around line 184:
    https://github.com/beurtschipper/Depix/blob/f2353b6f5e2334ae61182aceda32d486eb92b9ae/depixlib/functions.py#L184-L194
  • Assume that relative to singleResult, its ABOVE_PIXEL and BELOW_PIXEL both have only one "valid match".
  • By saying "valid match" we mean that the counter totalMatches in line 214 would be increased.
  • After the loops, when it comes to the check if totalMatches == 1:, it would fail.
  • However, the two pixels have indeed been determined by this singleResult, and should be taken into account.

Modification

• dataSeen: Changed from list to set for better performance.
• Instead of a single integer counter, I used a dictionary matchCount to keep track of the counter for each pixel.
• I changed the variable name xDistanceRectangles -> xDistance, but change it back if you think that'll be better.
• I guess line 193-195 are the main source of time decrease.

Before

After