Release 3.3.0 (2025/08/18)

[fabiencastan]

Highlights

New Pipelines

• Color Calibration: Automated color correction from color charts
• Raw to EXR conversion: Professional image format processing
• Object Reconstruction: Targeted reconstruction with automatic object segmentation
• Turntable Object Reconstruction: Streamlined workflow for rotating object capture
• 360° Object Reconstruction: Reconstruction of complete dual-sided scanning
• LiDAR Processing: Native E57 file import with integrated mesh generation
• Multi-View Photometric Stereo: Advanced surface detail reconstruction with multiple light sources for each viewpoint.

Pipelines Improvements

• Camera Tracking pipeline: improved stability and reliability
• Introduced experimental fine-grained pipelines for increased modularity and workflow flexibility

Core Enhancements

• Python Bindings Integration: Enhanced AliceVision accessibility with native Python support for streamlined Machine Learning workflows

Major Features

• Add initial Python binding with SWIG PR
• First Lidar Meshing Application PR
• Multiple shots: Align and merge multiple SfM from feature matches PR
• [meshroom] Add color calibration pipeline PR
• [pipeline] New photogrammetry of the two sides of an object PR
• [pipeline] New pipeline for object reconstruction from a turntable PR
• [pipeline] New pipeline for the photogrammetry of an object using AI semantic segmentation PR
• Add E57 importer PR
• Add rawImageConversion pipeline PR
• [meshroom] Add new "Lidar Meshing" template PR

Features

• ApplyCalibration: Lens Calibration is done on Open Gate plates, but camera tracking could be done on cropped plates PR
• [software] new MeshRemoveUnseenFaces software PR
• [core] Separate Camera and Distortion Types in Model Naming PR
• PLY loader to SfMData landmarks PR
• Lidar decimation and merging PR
• Add New Undistortion Models PR
• Undistortion Calibration PR
• [Photometric Stereo] MultiView fusion in Texturing PR
• Colorizing node PR
• [sfm] New sfmFilter software: split sfm based on filenames regex PR
• [export] Add export of animated distortion in Nuke PR
• Use lidar as constraints in sfm PR
• Enable merging of multiple sfms PR
• [build] SwigBinding: Add support for the Python binding with SWIG on Windows PR
• New tool to import external Alembic files with an animated camera PR
• Add masks to export images PR
• features detector checks for mask in a region PR
• Add numpy to python bindings PR
• Add module image to python bindings PR
• RelativePoseEstimating accepts an image pairs list to reduce complexity PR
• SwigBinding: Add Python binding for matchingImageCollection IO functions PR
• [sfm] Add new parameters to the SfM process PR
• [doc] README.md: Add DeepWiki link, the AI documentation you can talk to PR

Other Improvements

• Start Development 3.3.0 PR
• Add estimator states to sfmdata PR
• Update Color Space Extraction from image Metadata PR
• ImageSegmentation: add an option to choose between cpu and gpu PR
• Fix validity check on distortion object PR
• Enhance controllability and readability by ensuring non direct access to data members PR
• Export lens grid undistorted PR
• Pixel Aspect Ratio application PR
• Add Bundle Adjustment unit tests PR
• Remove specifics required by Eigen prior to C++17 PR
• Remove Boost-Filesystem and replace it with the std::filesystem PR
• Clean-up .clang-tidy and apply camelBack case on functions and variables for some modules PR
• Add option to keep input filename in imageSegmentation PR
• [software] Disable clang-format around program options PR
• [software] Apply clang-formatting on all the executables PR
• FeatureExtraction: Add analytical fisheye mask PR
• [sfmData] Rename functions and files to respect case rules PR
• Remove all remaining references to Boost-Filesystem PR
• Add aarch64 support (adjust PR1616) PR
• Add camera color spaces PR
• Rename Python aliceVision module into pyalicevision PR
• E57 importer is now generating multiple sfmData PR
• Clean and rework meshing code PR
• [sfm] Handle specific alignment case for pure translations PR
• [sfm] Sequential: Update default value for Ransac's max iterations PR
• [keyframe] Set default values for some uninitialized variables PR
• [aliceVision] Minor bugfixes across the project PR
• SwigBinding: Extend the binding of camera and add initial binding for other modules PR
• Option to Remove absolute paths to linked libraries PR
• [software] New SelectConnectedViews software PR
• [software] KeyframeSelection: Update default minimum number of keyframes PR
• [mesh] visibility from mesh itself: check point is in camera frustum PR
• Catch error codes thrown by std::filesystem::exists PR
• Add missing standard header include PR
• Export radial3 distortion to nuke PR
• [pipeline] FeatureMatching: Update default value for Ransac's maximum number of iterations PR
• [Cmake] split sfm library PR
• [sfm] update tracks PR
• Solve a set of bugs on cameraInit PR
• anisotropic bounds for sfm PR
• Add SfM Expanding application PR
• ApplyCalibration: Add option to use a Lens calibration file from 3dequalizer in json PR
• Use focal length from json and fix typos in parameters PR
• Add locks for intrinsics PR
• sfmTransform: Add option to lineup camera motion with object/lidar given an external camera pose PR
• Distortion export for pipelines without lens grids PR
• [hdr] Upgrade on HDR brackets estimation PR
• Clean up focal length and change anamorphic model PR
• [sfmData] ImageInfo: Add support for ARRI- and DJI-specific focal length metadata PR
• Depth/normal maps rendering from Mesh PR
• Remove imageCache code from alicevision PR
• [utils] Pose injection from JSON file PR
• Merge symbolic and automatic SfM PR
• Add global rotation estimation app PR
• Add SWIG binding of camera shared pointers PR
• Adding logs to sfmExpanding PR
• [docker] Fix export of ALICEVISION_* environment variables PR
• Build OpenCV with FFmpeg on Linux PR
• [docker] Update CUDA minor version to 12.1.1 PR
• Clarify check-cpu.sh documentation PR
• Update sfmBootstraping parameters PR
• Add Equirectangular camera support for SFM PR
• sfmMerge can merge Intrinsics PR
• Mask processing app PR
• Fix typos in README and PR template PR
• Output second sfm for visualisation PR
• Fix typos in src/aliceVision PR
• Reduce intrication of distortion with intrinsics PR
• Adding logs to export USD PR
• Fix typos in src/software PR
• Fix source typo in software/pipeline/main_lightingCalibration.cpp PR
• Fix various remaining typos PR
• fix sfmMerge PR
• Versatile image exporting PR
• Fix error on generated depthmap EXR PR
• Upgrade compilation options and CI for C++20 support PR
• Multiple lineup at different frames PR
• Fix some remaining typos PR
• Remove unused "ci" folder PR
• [sfmDataIO] Fix Alembic dependency requirement for unit test PR
• Survey Injection PR
• [meshroom] Add AliceVision nodes and templates for Meshroom PR
• Reduce compilation warnings and clean-up files PR
• [meshroom] ExtractMetadata: Move pyalicevision import into the processChunk method PR
• [docker] Copy meshroom folder to be able to install nodes and templates PR
• Color checker detection and correction update PR
• Fix division zero at Texturing PR
• Add TracksMerging node and partially add the tracks module to the Python binding PR
• Deep merging of multiple tracks sets without using original matches PR
• [utils] Image Processing: Add simultaneous support for regex and input folders PR
• Add application to generate virtual tracks PR
• Upgrade experimental pipelines for better integration in Meshroom PR
• [docker] Ubuntu: Use CMake from standard repo instead of Kitware's PR
• [meshroom] CamTrack: Provide PrepareDenseScene with segmentation masks PR
• [sfmData] Add focal length naming convention PR
• Global binding PR
• [utils] Optimize exportImages application when only undistort is needed PR
• [meshroom] SketchfabUpload: Import requests at the function-level PR
• Update image module of Python binding PR
• SwigBinding: Add binding for the feature and matching modules PR
• [meshroom] Use CamelCase for all attribute labels PR
• Update python test with oiio metadata writing PR
• Add more logging to new pipeline PR
• Output modified sfmData in exportImages PR
• Force Pixel Aspect Ratio in Distortion Calibration PR
• Improve SWIG Module Type Sharing for Enhanced Interoperability PR
• [blender] Fix .obj import for blender 4 PR
• Improving python bindings for image and feature modules PR
• SwigBinding: Add support of Landmarks PR
• SwigBinding: Expose ProgressDisplay PR
• Add parallelization tool to optimize chunk computation distribution PR
• [meshroom] Fix missing connection in "CameraTracking without calibration experimental" pipeline PR
• [meshroom] ImportAlembic: Add missing period in file extensions PR
• [meshroom] ScenePreview: Hide backfaces in 3D render PR
• [meshroom] MultiViewPhotometricStereo: Add missing Publish node PR
• [nodes] Blender: Rescale rendered images based on their pixel aspect ratio PR
• [nodes] ExportDistortion: Option to export lens grid undistorted PR
• [nodes] ImageProcessing: New option to apply the Pixel Aspect Ratio PR
• [pipelines] Enable Keep Filename for all ImageSegmentation nodes in templates PR
• [pipelines] Panorama: Enable "Minimal 2D Motion" for FeatureMatching PR
• [pipelines] PanoramaFisheyeHdr: Run FeatureExtraction after PanoramaInit PR
• [pipelines] PhotogAndCamTrack: Update matching mode to "Exhaustive" PR
• [nodes] ImageSegmentation: Support loading masks in 2D Viewer if they are named after the input images PR
• [nodes] CameraInit: check allowedCameraModels validity PR
• [nodes] Use utilitary constants for common ChoiceParams lists of values PR
• [nodes] ExportAnimatedCameras: rename as ST maps (instead of UV) PR
• [pipelines] NodalTracking: Apply masks in ScenePreview PR
• [nodes] New node MeshRemoveUnseenFaces PR
• [nodes] StructureFromMotion: Update default value for the maximum number of Ransac iterations PR
• [nodes] CameraInit: Separate Camera and Distortion Types in Model Naming PR
• [nodes] Update the ressources needs for the SFM node PR
• [nodes] New SelectConnectedViews node PR
• [nodes] KeyframeSelection: Update default minimum number of keyframes PR
• [nodes] ExportanimatedCamera: change default value for the principal point correction PR
• [nodes] FeatureMatching: Update default value for the maximum number of Ransac iterations PR
• [nodes] ImageProcessing : use keepImageFilename attribute PR
• [nodes] Fix outputImages if no export in ExportAnimatedCamera PR
• [nodes] Blender scripts: Add missing __init__.py PR
• [nodes] SphereDetection: Increase max value for sphereRadius PR
• [nodes] Use principal point in scenePreview PR
• [nodes] ApplyCalibration: Add option to use a Lens calibration file from 3dequalizer in json PR
• [pipelines] Add distortion export for templates without calibration PR
• [nodes] Use bindings for HDR brackets estimation PR
• [nodes] LightingCalibration: Exclude visualization output from command line PR
• [nodes] LdrToHdr: Move import of pyalicevision within functions PR
• [nodes] Add new SfMChecking node PR
• [pipelines] PhotogCamTrack: Invert masks for segmentation PR
• [nodes] Add Global Rotation Estimating node PR
• [nodes] ExportDistortion: Add animated distortion option PR
• [nodes] Fixed spelling mistake in FeatureExtraction.py PR
• [pipelines] Experimental: Align and set colors for all nodes PR
• [nodes] Add nodes for versatile image exporting PR
• [nodes] Replace desc.Node.internalFolder with {nodeCacheFolder} PR
• [pipelines] Nodal Camera Tracking: Add a new template with the calibration of the distortion PR
• [nodes] Add new SfMPoseInjecting node PR
• [meshroom] Fix ambient typo on MVPS template PR

Bugfixes

• [image] io: Rely on OIIO to determine if an image file exists or not PR
• Fix typo in bundle adjustment unit test PR
• [docker] Fix the build of Docker images for Ubuntu PR
• EXR STMaps use floating point precision PR
• Add fstream include to import e57 PR
• [release] fix script to check dependencies PR
• ExportAnimatedCamera export alembic with complete cameras information PR
• [sfmData] fix relative to absolute path conversion for features/matches folders PR
• fix camera pose lock I/O PR
• Fix pipeline when LdrToHdrMerge is used on single brackets PR
• Fix crash on convert distortion PR
• Minor checks and log PR
• [image] Fix erroneous color space detection for images that contain "raw" in their filepath PR
• Fix SfM when using undistortions PR
• Add missing fstream include in ApplyCalibration PR
• [software] PanoramaSeams: Check that the input is not empty PR
• Inverted flag check in intrinsics PR
• fix alembic I/O PR
• [utils] Fix applyCalibration logic for isDesqueezed PR
• [pipeline] sfmBootstraping: Fix inversion between min and max PR
• [FuseCut] Fix: Check cam vertex validity in meshing PR
• Add standard include for fstream PR
• [segmentation] Image segmentation CPU bugfix and ONNX output update PR
• fixed bug in sfmExpanding PR
• SfmBootstrapping: Fix bug when parallax angles are less than 1 degree PR
• Update minimal required CMake version to 3.5 for all modules PR
• [src] Fix dependencies depending on build options PR
• [track] Use std::map instead of std::unordered_map to represent Tracks PR
• Update SWIG code to remove errors PR
• [docker] Dependencies: Add a symlink to python3 to make python callable PR
• [image] Remove static keyword from isSupported and getSupportedExtensions PR
• [fix] Remove invalid default value in ldr2hdrCalibration PR
• [Dependencies] Adapt code to new Eigen PR
• Fix bug in sfmMerge PR
• [mesh] Fix mesh appearing whitened in 3D viewer PR
• [MVS] meshPostProcessing: bugfix in the cleanup of removed free points PR
• [sfm] statistics: fix statistics to avoid troubles in particular cases PR
• Update experimental pipelines PR
• Remove hack with Custom camera model PR
• [pipeline] Adding missing link in photogrammetry pipeline PR
• [Fix] Initialize resectionid depending on previous sfm PR
• Update json utility functions PR
• Remove error or warning for some compilers PR
• Fix nodal pipeline including scenePreview crash PR
• Update the solver's parameters for more robust unit tests PR
• [meshroom] PhotogrammetryObjectTwoSides: Remove hard-coded paths to models PR

CI, Build and Documentation

• [doc] Fix RTD build + other fixes/improvement of the doc PR
• [cmake] AllInOne: update to suitesparse-7.3, ceres-2.2 PR
• [ci] Update link to the VCPKG archive for the Windows CI PR
• [docker] Fix the build of Docker images for Ubuntu PR
• [cmake] Dependencies: Update zlib version to 1.3.1 PR
• [build] Add libe57format as an internal dependency PR
• [build] Add libe57format as an internal dependency PR
• [build] Update the list of installed features for OpenImageIO with VCPKG PR
• [build] Update Geogram to 1.8.8 PR
• Fix coverage in cmake PR
• Update boost and oiio dependencies PR
• Remove OpenMesh from internal dependencies PR
• Add condition to check dependency MeshSDFilter folder PR
• libE57Format in Copying.md PR
• Externalize nanoflann PR
• MeshSDFilter: Remove submodule PR
• [ci] Windows: Add appropriate flags to compile with CUDA < 12.1 PR
• [build] Fix some C++17 incompabilities PR
• [build] Update SWIG version to 4.3.0 PR
• CI: Move from CentOS to Ubuntu for Linux PR
• [docker] Add Dockerfiles for Rocky 9 PR
• [cmake] Update download link for Boost 1.84.0 PR
• [src] Comply with nanoflann >= 1.6.0 changes PR
• [ci] Windows: Only install necessary components from the CUDA Toolkit PR
• [ci] Windows: Update vcpkg archive with latest dependencies versions PR
• [ci] Windows: Update vcpkg archive with SWIG support PR
• [ci] Update the path to the segmentation nodes for Meshroom in the CI PR
• Update minimal required CMake version to 3.5 for all modules PR
• [src] Fix dependencies depending on build options PR
• [track] Use std::map instead of std::unordered_map to represent Tracks PR
• Update SWIG code to remove errors PR
• [docker] Dependencies: Add a symlink to python3 to make python callable PR
• Dependencies update PR
• [src] Update DLL discovery in __init__.py for the binding on Windows PR
• Update Dockerfiles PR
• [docker] rocky: update gxx and enable cctag PR
• [ci] Windows: Update vcpkg archive with 2025.07.08 tag PR
• [build] Enable the build of OpenCV's and OpenImageIO's Python bindings PR
• [doc] Update INSTALL.md PR
• [doc] README.md: Add DeepWiki link, the AI documentation you can talk to PR
• [cmake] Dependencies: Update OpenCV to 4.12.0 PR

Updates to the Sensor Database

• Add ASUS_I01WD (Asus ZenFone 6) to sensorDB PR
• Adding Google Pixel 6a to cameraSensors.db PR
• Update cameraSensors.db PR
• Update cameraSensors.db PR
• Update cameraSensors.db PR
• Update Fairphone cameraSensor PR
• Fairphone 5 cameraSensors.db fix PR
• [cameraSensors] OMNIVISION uctronics and gophotonics PR
• [cameraSensors] fix Samsung SM-G998B PR
• [cameraSensors] Add DJI Mini 2 SE PR
• Added insta360 x3 sensor to db PR
• [sensordb] Fix typo of 'OnePlus' devices PR
• [sensordb] OnePlus 8 Pro PR
• Fix Sonx RX0 II camera model PR
• Update cameraSensors.db Sony ILCE-6700 PR
• Update cameraSensors.db PR
• Update cameraSensors.db DJI Mini3Pro PR
• Update cameraSensors.db with Pixel 8 and Pixel 8 Pro PR
• Update cameraSensors.db PR

Contributors

ahdistefano, albertoporri, Alxiice, AncoraIo36, andrewgreen5610, apirek, Armado001, balloonalpha, c-ewing, Cal-Bur, cbentejac, CedricThebault, demoulinv, Erol444, EternalUnderperformer, fabiencastan, heyvian, ICIbrahim, JackBoosY, jmelou, luzpaz, maggo1404, male4a, mh0g, natowi, NicholasFlamy, paweller, predat, servantftransperfect, simogasp, theremygit, woutkolkman

---

This discussion was created from the release Release 3.3.0 (2025/08/18).