E.g., how to preview the search results on Windows using xnview. Or, make it easier to do this.

Context: https://www.youtube.com/watch?v=tAJHXOkHidw&lc=UgxPjMMxOEehG21-osd4AaABAg

I tested that searching text in images. It works, though not so good. Any idea to improve the performance of searching text in images? Thank you.

Hello!

I have noticed that rclip saves the database based on an environment variable or a system data directory:

Is there any reason why we can't configure rclip to save the database on the current directory ala git? If I am not mistaken, that would allow users to move/copy these folders and keep a database of the cached images. This would be really useful if someone managed their albums through a cloud platform, allowing them to also upload the database file and share it across devices.

The installed mintty can be called "rclip terminal" (similar to what git installer does by calling it "Git Bash" on Windows).

If you want to increase the priority of this Issue, comment and give it a thumbs up.

Context: https://www.youtube.com/watch?v=tAJHXOkHidw&lc=UgxPjMMxOEehG21-osd4AaABAg.9uZ_CVQputM9vMaC3xTFig.

I'm getting 3.7 seconds average execution time on my laptop (i7-7700HQ) when searching through 73 thousand images, where:

0.85s is "import clip"
1.75s is model loading
0.47s is querying the data from the SQLite DB
0.26s is actual querying

On my NAS (Intel Celeron J3455), it executes in 7.8s average, where:
1.73s "import clip"
3.22s model loading
1.87s getting features
0.8s search

~65% of the execution time is the CLIP model loading, so reducing this time can be the vector worth exploring first.

One of the options to take is to make rclip a daemon that has the model loaded all the time and query it, but this results in the extra RAM consumption of a few hundred megabytes, and, given that querying is a relatively rare operation, I don't like it.

Another option is to convert the CLIP model to Tensorflow Lite. It can be tricky but should reduce both the RAM consumption and the execution time, so it may be worth exploring.

UPD: this feature is being worked on and will be released as a separate tool, rtag. You can follow the development by starring and watching the rtag repository. This Issue will be closed once rtag v1 is released.

Tags should be taken from a fixed, pre-defined set. The ImageNet competition tag list is a good place to start, wikientities is another one.

The v1 of the feature should look like this from the user perspective:

$ rclip tag
This operation will overwrite the tags stored in your images' EXIF metadata. This will change the image files and will make you loose any existing keywords stored on your images.
Proceed? y/n y
/photos [#####]

For v1, it should write the tags from the list of the ImageNet-1k tags, based on a certain threshold (should be determined empirically).

Interested user: https://www.youtube.com/watch?v=MsTgYdOpgcQ&lc=Ugzl8qVAiVsek_2TwoZ4AaABAg

when there are unicode file names, rclip runs well only in the console window. However, if you redirect the output to file, pipe, etc. it will throw errors. This blocks me from integrating rclip into my app.

See below screenshot:

• 1st run: [OK] console output
• 2nd run: [FAILED] redirecting output

A good example of what I mean is http://image-search.0ape.com/search?q=naked%20-person.

Thanks, @ramayer for the rclip-server and the inspiration!

Requirements:

• the current interface rclip "query" should continue to work
• it should correctly support following examples:
  rclip "multi word query -word1 -word2"
rclip "query -'multi word negative'"
rclip 'query -"multi word negative"' (i.e., it should work with any type of quotes)

rclip "multi word query -word2 some other words" should error with the suggestion to use quotes around the negative since it's unclear how to parse it

Ensure that instead of multiple progress bars, rclip displays a single progress bar which will include the total number of files as soon as possible

https://snapcraft.io/

It's low priority because we already build AppImage for Linux.

• ARW
• DNG
• CRW
• CR2
• CR3

We should use iTerm inline image protocol, which is supported by:

• iTerm2 on macOS
• Konsole on Linux (doesn't work on Ubuntu; maybe need to get the latest Konsole https://invent.kde.org/utilities/konsole/-/merge_requests/594)
• wezterm on Linux

We may also add support for sixel graphics for terminals like xterm.

Comment if you want to see this feature implemented.

pip version 23.2.1

$ pip install rclip
$ rclip cat
---
checking images in the current directory for changes; use "--no-indexing" to skip this if no images were added, changed, or removed
0images [00:00, ?images/s]

100%|███████████████████████████████████████| 354M/354M [00:41<00:00, 8.58MiB/s]
score	filepath

No results.

Sometimes it's more convenient to search for "a picture similar to the one I have in mind", instead of trying to describe it with words. Especially considering the extremely short phrases CLIP allows.

I'd like to be able to run:

 rclip https://i.redd.it/a7jf5rlghsz51.jpg

to return pictures of similar buildings like these.

It would also be nice to be able to search for pictures similar to a local file, like rclip /tmp/turtle.jpg and/or rclip file:///tmp/turtle.jpg.

This branch has a working work-in-progress for both of those features; but probably isn't ready for a pull request yet because it has a python3.9 dependency (str.removeprefix is a 3.9 feature). If you're interested, I'll remove that dependency and open a pull request.

This is a low priority because we already support the iTerm Inline Graphics protocol, which allows you to preview images in iTerm2, Konsole 22.04+, wezterm, Mintty, and mlterm.

Vote and comment on this issue if you want to prioritize implementing the support for Sixel and kitty protocols. And, if you feel like it, give a stab at it yourself and submit the PR. Contributors are welcome! :)

Things to test first:

• handling of empty directories
• handling of nested directories
• handling of multiple directories
• addition of the images
• removal of the images

In the recent ML news, I noticed mentions of FasterViT, a new model with some architecture changes that is said to outperform others in the performance to accuracy ratio. I immediately wondered if it can give tools like this any performance gain.

https://github.com/NVlabs/FasterViT

Changelog for pre-release versions should include all of the commit messages for commits made between the new and the previous version tag, in the same way, they do now.

Changelog for release versions should include only ^(feat)|(fix)|(perf)|(BREAKING CHANGE) commit messages for commits made between the new and the previous release version tag.

We need to build a self-contained executable for Windows and make GHA build and release it alongside Linux's AppImage.

Consider using either winget or chocolatey or scoop. Or is there anything better?

Automate homebrew tap updates and add rclip@beta Formula for pre-releases.

https://github.com/yurijmikhalevich/homebrew-tap

E.g., like rsync has the --exclude=PATTERN that can be repeated...

I see this in main.py:

EXCLUDE_DIRS = ['@eaDir', 'node_modules', '.git']
EXCLUDE_DIR_REGEX = re.compile(r'^.+\/(' + '|'.join(re.escape(dir) for dir in EXCLUDE_DIRS) + r')(\/.+)?$')

Do you have any interest in having additional exclude directories specified as command line arguments? Or perhaps an e.g. .rclipskip, .rclipexclude, .rclipignore file that contains directory patterns?

https://github.com/codesee-io/oss-port
https://www.oss-port.com/

For those who wants extra preciseness and ready to sacrifice speed for them:

Stats from a dataset of 269 photos on a much slower NAS (using the v1.2.3 and v2.0.0-pre AppImage distributions):

• ingestion: 3m56.626s vs 125m0.507s (+3070%)
• search: 0m18.064s vs 3m19.742s (+1006%)

More details are here.

Like and comment if you want to see the largest CLIP model in rclip.

Homebrew 4.2.12

$ brew install yurijmikhalevich/tap/rclip
$ rclip "cat"
---
Traceback (most recent call last):
  File "/opt/homebrew/bin/rclip", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/opt/homebrew/Cellar/rclip/1.7.25/libexec/lib/python3.11/site-packages/rclip/main.py", line 205, in main
    arg_parser = helpers.init_arg_parser()
                 ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/rclip/1.7.25/libexec/lib/python3.11/site-packages/rclip/utils/helpers.py", line 144, in init_arg_parser
    if is_mps_available():
       ^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/rclip/1.7.25/libexec/lib/python3.11/site-packages/rclip/utils/helpers.py", line 153, in is_mps_available
    import torch.backends.mps
ModuleNotFoundError: No module named 'torch'

Compute and store image hashes, and add logic to handle image renaming properly.

Is there a way to run rclip without a query, just to update the index? Use case would be to have it run as a cron job so when a query needs to be run, there wouldn't be any waiting for indexing.

Write a formula to install rclip with brew on macOS.

It's a good idea to rebuild the installer as an MSI to streamline the installation process first.