Metadata-Version: 2.1
Name: clorps
Version: 0.1.0
Summary: CLORPS: A module for CLIP, LPIPS, and ORB based image similarity.
Home-page: https://github.com/vaidatascientist
Author: Vaibhav Gupta
Author-email: vaibhavgupta.ggwp@gmail.com
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.6
Description-Content-Type: text/markdown

# CLORPS

**CLORPS** is a Python package for calculating image similarity based on three complementary techniques: **CLIP embeddings**, **LPIPS (Learned Perceptual Image Patch Similarity)**, and **ORB (Oriented FAST and Rotated BRIEF)**. It provides a combined similarity score for one image compared to another image or a set of target images. Ideal for image retrieval, similarity-based ranking, and image comparison tasks.

## Features

- **CLIP Embeddings**: Uses OpenAI’s CLIP model for embedding-based similarity.
- **LPIPS Similarity**: Calculates perceptual similarity using LPIPS.
- **ORB Keypoint Matching**: Traditional ORB-based similarity for structural comparison.
- **Combined Score**: Normalizes and combines the three scores for a final similarity metric.

## Installation

```bash
pip install clorps
```
**Usage**
```bash
from clorps import CLORPS

# Initialize CLORPS instance
clorps_instance = CLORPS()

# Paths to input and target images
input_image_path = "/path/to/input/image.jpg"
target_image_paths = ["/path/to/target/image1.jpg", "/path/to/target/image2.jpg"]

# Calculate combined similarity scores
combined_scores = clorps_instance.calculate_combined_similarity(input_image_path, target_image_paths)
print("Combined similarity scores:", combined_scores)
```
**Input:** Path to the input image and either a single target image path or a list of target image paths.
**Output:** A list of similarity scores, one for each target image.

**Requirements:**
- Python 3.6+
- torch
- open_clip_torch
- lpips
- numpy
- scikit-learn
- Pillow
- opencv-python

**LICENSE**
This project is licensed under the MIT License.
