The goal of this project is to create a image comparator, to help find duplicates and similar images in an image folder.
At the moment, it is using the ORB feature extractor with a Ransac algorithm to detect similar images. It will output a sorted list of possible matches. The threshold is still fairly low during development, to maintain a high recall.
Just type
./comparator/comparator.py <target> <folder>
./comparator/orb.py <target> <folder>target is the file we want to compare, and folder is where we want to look for duplicates. It is not recursive yet.
It will then print a sorted list of the matched images, with the best matching first. A higher score means a higher match.
The syntax is similar
./interface.py <target> <folder>It will display the results in the following way:
- Support recursion in folders
- Use a more robust framework for the script parameters
- Find a way to use FLANN in a multi-process way for knn matching
- Handle mirrored images
- Tweak thresholds when matching features
- Handle case for images without features
- Save features in a file to speedup mutltiple searchs
- Profile the script
- Find a usable feature extractor to improve performances
- ORB is used at the moment.
- Started with SIFT, but really really slow during generation and matching.
- Display score in GUI
- Add option for the number of nearest neighbors we want to have
- The algorithm can now make a decision, so not useful anymore