SIFT

Implementation of scale-invariant feature transform(SIFT)

Description

This repository contains implemetation of the SIFT:

1. Keypoint Detector
2. Descriptor

It is done by using only numpy features and opencv read/write methods.

Keypoint detector is a method, which implements several stages of finding extremum points and their filtering.
Filters:

1. Contrast filter;
2. Shift filter;
3. Edge filter.

Environment

To setup environment, please install required python packages:

pip install -r requirements.txt

Run example:

You can test this repository by compute one image:

python3 main.py

or to evaluage this pipeline with the different datasets to get their accuracy:

python3 evaluation.py