ImageLookup

This is a fast algorithm that has the following usecase:

• you have an image or a patch
• you have tons of other images (like a database, maybe millions of images)
• run the algo and it will tell you where did your image or patch come from

How to run it?

• place your image database (images) in sources
• place your query images in queries
• run the following:

python image_lookup.py

• enjoy!

Dependencies

numpy, tensorflow, opencv, tqdm, sklearn

you can of course check other parameters, which are defined in the argparsed in utils.py

Evaluation

the script will print out an accuracy score if a gt.txt file is present in the main dir. The gt.txt is a simple text file that looks like this:

query source
query_imageX source_imageY
...

where query_imageX and source_imageY are the image names for the source and query (without extension)

WIP