simikit 0.1.3

Image Similarity Toolkit in Python

SimiKit: Image Similarity Toolkit in Python

English | 中文

Overview

SimiKit is a toolkit for commonly used image similarity algorithms. This project provides various tools to help developers quickly compare the effects of multiple image similarity algorithms, and assist developers in selecting an image similarity algorithm that best meets their needs.

Installation

1. Install by pip

pip install simikit

2. Install by uv

• 2.1. Install uv(if not currently installed)

curl -LsSf https://astral.sh/uv/install.sh | sh  # macOS and Linux
# on Windows:
# powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

• 2.2. Clone the repository:

git clone https://github.com/yie1d/simikit.git  
cd simikit

• 2.3. Create a new virtual environment and activate it:

uv venv
source .venv/bin/activate  # On Unix/macOS
# Or on Windows:
# .venv\Scripts\activate

• 2.4. Install dependencies:

uv sync

Basic Usage

1. extract image features

from simikit.features import AHash, Vit, DinoV2

print(DinoV2().encode('./t1.png'))
print(Vit().encode('./t1.png'))
print(AHash().encode('./t1.png'))

2. use comparator by multiple algorithms

from simikit.api import Comparator
from simikit.features import AHash, DHash
from simikit.metrics import hamming_distance

comparator = Comparator([
    (DHash(16, vertical=True), hamming_distance),
    (AHash(16), hamming_distance),
    (AHash(8), hamming_distance),
])

print(comparator.compare_image(
    './t1.png',
    './t2.png',
))

Supported Algorithms

• HASH
  • Average hashing
  • Difference hashing
  • Perceptual hashing
  • Wavelet hashing
• Transformer
  • VIT
  • DINOv2

Contribution

Thank you for your interest in simikit. Submissions in all aspects are welcome. Let's work together to make simikit better!

Future Plans

• Add more image similarity algorithms

If there is any similarity algorithm that you want but is not currently available in the simikit, you are welcome to raise it in the Issues section!