---
id: imagededup
version: "0.3.3.post2"
license: Apache-2.0
license_treatment: permissive
maintenance: aging
---

# imagededup — Package for image deduplication

License: permissive · Maintenance: aging · Downloads: 101.1K/mo

## What it is and what it does

imagededup is a Python package for identifying duplicate and near-duplicate images within collections. It provides multiple detection methods: hashing algorithms (PHash, DHash, WHash, AHash) for exact duplicates, and convolutional neural networks for finding near duplicates even when images have been transformed. You can generate encodings for all images in a directory, find duplicates using those encodings, and visualize results.

The package depends on torch, torchvision, Pillow, scikit-learn, PyWavelets, tqdm, and matplotlib. It runs on Linux, macOS, and Windows with Python 3.9 or later. An evaluation framework is included to assess deduplication quality against ground truth. The last release was 364 days ago and maintenance is aging, so updates may be infrequent.

Use it for:

- Clean up large photo libraries by identifying and removing visually identical or very similar images.
- Detect near-duplicate images in datasets that have undergone minor transformations like compression or resizing.
- Evaluate the effectiveness of image deduplication methods using provided benchmarking and evaluation tools.
- Build custom deduplication pipelines by selecting from multiple hashing algorithms or supplying your own CNN model.
- Batch process image directories to generate perceptual encodings for downstream similarity analysis.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Finds exact and near-duplicate images in collections using hashing algorithms and convolutional neural networks, with support for custom models and an evaluation framework.

Yes, if you need robust image deduplication and can tolerate the medium install friction from its deep learning dependencies. The package is stable (Production/Stable status), supports current Python versions, and has no known vulnerabilities. However, the aging maintenance status (364 days since last release) means updates and bug fixes may be slow; evaluate whether that risk is acceptable for your use case.

## Install

pip install imagededup
uv add imagededup
poetry add imagededup

## Installing imagededup

Before you install: Medium install friction due to 7 runtime dependencies including torch, torchvision, and scikit-learn. Prebuilt wheels available for Python 3.9–3.12 on Linux, macOS, and Windows. Last release was 364 days ago; maintenance status is aging.

License in practice: Licensed under Apache-2.0 (permissive), allowing commercial and private use with minimal restrictions. You may use, modify, and distribute the package freely provided you include the license notice.

Quickstart:

pip install imagededup

from imagededup.methods import PHash
phasher = PHash()
encodings = phasher.encode_images(image_dir='path/to/images')
duplicates = phasher.find_duplicates(encoding_map=encodings)

Requires Python 3.9 or later. torch and torchvision are heavy dependencies; initial installation and model loading may take significant time and disk space.

Verify before relying:

- Whether the package's CNN models are pre-trained or require separate download/training steps.
- Memory requirements for processing large image collections.
- Performance characteristics of each hashing method on modern hardware.
- Current download volume and active user base.

## Package facts

- License: Apache-2.0 (permissive)
- Python support: supports_current
- Install friction: medium
- Maintenance: aging
- Downloads: 101.1K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

find duplicate images, image deduplication, near duplicate detection, perceptual image hashing, CNN image similarity, batch image comparison, image dataset cleanup, image-processing, duplicate-detection, computer-vision

[View on SkillFed](https://skillfed.io/packages/imagededup) · [View on PyPI](https://pypi.org/project/imagededup/)