memery 0.0.2

search your image collections with natural language

memery

Search over large image datasets with natural language and computer vision!

Install

The necessary CLIP and torch packages will be installed by pip. You might want to make sure you have a sane CUDA environment before and after this step if you're trying to use GPU. If you don't have a GPU, memery should still work on your CPU.

If you have any trouble please open an issue! I want to make this package useful for as many people as possible. Help me know what's going wrong :)

pip install memery

How to use

Use GUI

The browser GUI is a Voilá app, which basically runs a Jupyter notebook without showing any of the code cells. To use this way, navigate to your memery folder in a terminal and run:

voila serve app.ipynb

If you're in a Jupyter environment, you can summon the GUI directly into an output cell like this:

from memery.gui import appPage

app = appPage()
display(app)

<memery.gui.appPage at 0x7f6c0e7c80d0>

Use CLI

From the command line, you can use memery on any folder and it will search for images recursively, returning a list object to stdout.

Pass the --n flag to control how many images are returned (default 10).

memery PATH/TO/IMAGE/FOLDER 'query' --n 20

Use as a library

Simply use queryFlow to search over a folder recursively! The folder will be indexed, if an index doesn't already exist. Then any new images will be CLIP-encoded, an Annoy treemap built, and a list of ranked filenames returned.

from memery.core import queryFlow
from memery.gui import get_grid

ranked = queryFlow('./images', 'dad joke')

Searching 78 images

get_grid(ranked, n=3)

print(ranked[:5])

['images/Wholesome-Meme-68.jpg', 'images/Wholesome-Meme-74.jpg', 'images/Wholesome-Meme-88.jpg', 'images/Wholesome-Meme-78.jpg', 'images/Wholesome-Meme-23.jpg']

TODO:

• Documentation and tests in notebooks
• Optimize GUI to cache results of searches
• Optimize the image loader and number of trees based on memory and db size
• Type annotations
• Search/browse by image or image+text
• Tags and filters
• Native GUI frontend client for various platforms
• Distributed server backend

DONE:

• Code for joining archived data to new data
• Code for saving indexes to archive
• Flows
• Cleanup repo
• CLI
• interactive GUI