PhotoDedup is a Windows desktop application for finding, reviewing, and safely resolving duplicate photo groups. It combines exact matching, perceptual similarity, optional AI-assisted comparison in the Full edition, Google Takeout metadata support, and a review-first PyQt6 interface.
| Lite interface | Full interface | Duplicate resolution |
|---|---|---|
 |
 |
 |
PhotoDedup ayuda a limpiar bibliotecas fotograficas grandes sin borrar a ciegas. Primero analiza los archivos, agrupa posibles duplicados y luego te permite decidir visualmente que foto conservar.
- Deteccion exacta por tamano/hash.
- Deteccion visual por hashes perceptuales.
- Analisis asistido por IA en
PhotoDedup-full.exe. - Edicion Lite sin dependencias pesadas de IA.
- Integracion con metadatos de Google Takeout (
*.json). - Revision por grupos con foto recomendada.
- Acciones seguras: mover duplicados o enviarlos a la papelera.
- Interfaz en ES / EN / PT.
- Cache local para acelerar analisis posteriores.
Usa la ultima version publicada:
https://github.com/wilkinbarban/photo-dedup/releases/latest
Artefactos principales:
| Archivo | Uso recomendado |
|---|---|
PhotoDedup-lite.exe |
Menor tamano, arranque rapido, flujo hash/visual sin IA. |
PhotoDedup-full.exe |
Analisis mas profundo con IA opcional cuando el runtime esta disponible. |
powershell -ExecutionPolicy Bypass -Command "iwr -UseBasicParsing https://raw.githubusercontent.com/wilkinbarban/photo-dedup/main/install.ps1 | iex"install.ps1 es el instalador unico. Si se ejecuta fuera del proyecto, descarga el repositorio oficial, valida la estructura, instala o actualiza la copia local y luego continua con la instalacion desde esa copia. Si se ejecuta dentro del proyecto, valida Python 3.14.x, prepara .venv, instala dependencias y abre PhotoDedup.
Para una copia ya descargada o clonada, ejecuta:
Iniciar.batpython -m venv .venv
.venv\Scripts\activate
python -m pip install --upgrade pip
pip install -r requirements.txt
python src\main\photo_dedup.pyPython compatible: 3.14.x; recomendado: usar el Python instalado en la PC y disponible en PATH.
Las dependencias de requirements.txt estan ajustadas para Python 3.14.x, incluyendo PyQt6 6.11, NumPy 2.4, OpenCV 4.13, PyTorch 2.12 y torchvision 0.27.
PhotoDedup intenta degradar de forma controlada:
- Si la IA no esta disponible, la app continua en modo hash/visual.
- Si detecta metadatos de Google Takeout, usa los JSON para enriquecer JPEG cuando es compatible, organiza fotos/videos por fecha y mueve los JSON procesados a
Json. - En carpetas Takeout parciales, tambien organiza medios sin JSON asociado: usa la fecha del nombre cuando existe y, si no puede inferirla, los mueve a
Sin_fecha. - Si un HEIC/HEIF no admite escritura EXIF directa, omite esa escritura sin detener la organizacion por fecha.
- Si una miniatura no se puede leer, se muestra un marcador visual neutro.
- Si el cache o la configuracion local estan corruptos, se usan valores seguros y se registra el detalle tecnico.
- Si una accion de mover/eliminar falla en algunos archivos, se informa el detalle por archivo sin ocultar los resultados exitosos.
Los ejecutables actuales no estan firmados con certificado de code signing. Windows puede mostrarlos como aplicacion de editor desconocido. Si confias en el origen oficial de GitHub Releases, puedes abrir manualmente Mas informacion y luego Ejecutar de todas formas.
PhotoDedup helps clean large photo libraries without blind deletion. It analyzes files, groups likely duplicates, and lets you visually decide which photo to keep.
- Exact duplicate detection by size/hash.
- Visual similarity detection through perceptual hashes.
- AI-assisted comparison in
PhotoDedup-full.exe. - Lite edition without heavyweight AI dependencies.
- Google Takeout metadata integration (
*.json). - Group-based duplicate review with a recommended keep choice.
- Safe actions: move duplicates or send them to the recycle bin.
- ES / EN / PT interface.
- Local cache for faster repeated scans.
Use the latest official release:
https://github.com/wilkinbarban/photo-dedup/releases/latest
Main artifacts:
| File | Recommended use |
|---|---|
PhotoDedup-lite.exe |
Smaller binary, faster startup, hash/visual workflow without AI. |
PhotoDedup-full.exe |
Deeper analysis with optional AI support when the runtime is available. |
powershell -ExecutionPolicy Bypass -Command "iwr -UseBasicParsing https://raw.githubusercontent.com/wilkinbarban/photo-dedup/main/install.ps1 | iex"install.ps1 is now the single installer. When executed outside the project, it downloads the official repository, validates the extracted structure, installs or updates the local copy, and continues from that local copy. When executed inside the project, it validates Python 3.14.x, prepares .venv, installs dependencies, and opens PhotoDedup.
For an already downloaded or cloned copy, run:
Iniciar.batpython -m venv .venv
.venv\Scripts\activate
python -m pip install --upgrade pip
pip install -r requirements.txt
python src\main\photo_dedup.pyCompatible Python: 3.14.x; recommended: use the Python installed on the PC and available in PATH.
requirements.txt is aligned with Python 3.14.x, including PyQt6 6.11, NumPy 2.4, OpenCV 4.13, PyTorch 2.12, and torchvision 0.27.
PhotoDedup is designed to degrade gracefully:
- If AI is unavailable, the app continues in hash/visual mode.
- If Google Takeout metadata is detected, JSON sidecars are used to enrich compatible JPEGs, organize photos/videos by date, and move processed JSON files to
Json. - In partial Takeout folders, media without a matching JSON sidecar is also organized: filename dates are used when available, otherwise files are moved to
Sin_fecha. - If HEIC/HEIF cannot be written through direct EXIF insertion, that write is skipped without blocking date-based organization.
- If a thumbnail cannot be read, the UI shows a neutral placeholder.
- If local cache or config files are corrupt, safe defaults are used and the technical detail is logged.
- If moving/deleting fails for some files, the app reports per-file details while preserving successful results.
Current executables are not code-signed. Windows may show them as unknown publisher apps. If you trust the official GitHub Releases source, open More info and then Run anyway.
PhotoDedup ajuda a limpar bibliotecas grandes de fotos sem exclusao cega. Ele analisa arquivos, agrupa duplicatas provaveis e permite decidir visualmente qual foto manter.
- Deteccao exata por tamanho/hash.
- Deteccao visual por hashes perceptuais.
- Comparacao assistida por IA em
PhotoDedup-full.exe. - Edicao Lite sem dependencias pesadas de IA.
- Integracao com metadados do Google Takeout (
*.json). - Revisao por grupos com foto recomendada.
- Acoes seguras: mover duplicatas ou enviar para a lixeira.
- Interface ES / EN / PT.
- Cache local para acelerar analises futuras.
Use a ultima release oficial:
https://github.com/wilkinbarban/photo-dedup/releases/latest
Artefatos principais:
| Arquivo | Uso recomendado |
|---|---|
PhotoDedup-lite.exe |
Binario menor, inicio mais rapido, fluxo hash/visual sem IA. |
PhotoDedup-full.exe |
Analise mais profunda com IA opcional quando o runtime esta disponivel. |
powershell -ExecutionPolicy Bypass -Command "iwr -UseBasicParsing https://raw.githubusercontent.com/wilkinbarban/photo-dedup/main/install.ps1 | iex"install.ps1 agora e o instalador unico. Quando executado fora do projeto, ele baixa o repositorio oficial, valida a estrutura extraida, instala ou atualiza a copia local e continua a partir dessa copia. Quando executado dentro do projeto, valida Python 3.14.x, prepara .venv, instala dependencias e abre o PhotoDedup.
Para uma copia ja baixada ou clonada, execute:
Iniciar.batpython -m venv .venv
.venv\Scripts\activate
python -m pip install --upgrade pip
pip install -r requirements.txt
python src\main\photo_dedup.pyPython compativel: 3.14.x; recomendado: usar o Python instalado no PC e disponivel no PATH.
requirements.txt esta alinhado com Python 3.14.x, incluindo PyQt6 6.11, NumPy 2.4, OpenCV 4.13, PyTorch 2.12 e torchvision 0.27.
PhotoDedup tenta degradar de forma controlada:
- Se a IA nao estiver disponivel, o app continua em modo hash/visual.
- Se metadados do Google Takeout forem detectados, os JSONs sao usados para enriquecer JPEGs compativeis, organizar fotos/videos por data e mover os JSONs processados para
Json. - Em pastas Takeout parciais, midias sem JSON associado tambem sao organizadas: datas no nome do arquivo sao usadas quando existem; caso contrario, os arquivos vao para
Sin_fecha. - Se HEIC/HEIF nao permitir escrita EXIF direta, essa escrita e ignorada sem bloquear a organizacao por data.
- Se uma miniatura nao puder ser lida, a interface mostra um marcador neutro.
- Se cache ou configuracao local estiverem corrompidos, valores seguros sao usados e o detalhe tecnico e registrado.
- Se mover/excluir falhar em alguns arquivos, o app mostra detalhes por arquivo sem ocultar os resultados bem-sucedidos.
Os executaveis atuais nao possuem assinatura de codigo. O Windows pode mostra-los como aplicativos de editor desconhecido. Se voce confia na origem oficial do GitHub Releases, abra Mais informacoes e depois Executar assim mesmo.
| Path | Purpose |
|---|---|
src/main/photo_dedup.py |
Canonical PyQt6 entry point. |
src/interfaces/ |
Main window, screens, reusable widgets, language dialog, and visual theme. |
src/modules/services/ |
Duplicate analysis, AI model, Takeout handling, and domain models. |
src/modules/config/ |
i18n, app state, cache, config, and history. |
src/modules/utils/ |
Logging, asset path resolution, and shared error helpers. |
requirements.txt |
Runtime dependencies pinned for Python 3.14.x. |
requirements-build.txt |
Build-only dependencies such as PyInstaller. |
install.ps1 |
Unified one-command installer and launcher for remote or local use. |
Iniciar.bat |
Local Windows launcher for manually cloned or downloaded copies. |
scripts/build_windows.ps1 |
Builds one Windows EXE flavor. |
scripts/build_variants.ps1 |
Builds Full and Lite variants. |
.github/workflows/ |
Release, build, and smoke-test automation. |
python -m py_compile src\main\photo_dedup.py src\interfaces\language_dialog.py src\interfaces\main_window.py src\interfaces\screens.py src\interfaces\theme.py src\interfaces\widgets.py src\modules\config\i18n.py src\modules\config\state.py src\modules\services\ai_model.py src\modules\services\analyzer.py src\modules\services\models.py src\modules\services\takeout.py src\modules\utils\errors.py src\modules\utils\logger.py src\modules\utils\paths.pyPackaging validation:
pip install -r requirements-build.txt
.\scripts\build_variants.ps1 -Version local-test -SmokeTestThis project is provided for educational purposes: PyQt6 desktop UI, image-analysis workflows, background processing, packaging, release automation, and resilient error handling. Use it only with photo libraries you own or are authorized to manage.
PhotoDedup is licensed under the GNU General Public License v3.0. See LICENSE.