Agisoft Metashape Pro 2.4.0: Photogrammetry & 3D Reconstruction Software

Agisoft Metashape Pro 2.4.0 is a pre-release version of the professional photogrammetry suite, published September 14, 2026 (build 23155). It introduces a new image matching algorithm, trainable point cloud classification, optimized texture blending, and native ARM64 support .

Important: This is a pre-release considered unstable and not recommended for production use. Project files created in 2.4.0 cannot be opened in Metashape 2.3 or earlier .

Key New Features

New Image Matching Algorithm

The algorithm for selecting key points and the descriptor have been updated. Matching now progresses from a coarse level to a detailed one, enabling more matches in complex scenarios such as fields, forested areas, and repetitive or uniform patterns .

  • Accuracy now defaults to Highest: The algorithm automatically selects features across all downscale levels (Lowest to Highest) for the most accurate alignment .

  • Guided Image Matching removed from the Align Photos dialog — its logic is now integral to the algorithm .

  • New parameters: Specify Key point limit minimum and Key point limit per Mpx (previously part of Guided matching). Detected points per image = (Key point limit per Mpx) × (image size in Mpx), but not lower than the minimum value .

Trainable Point Cloud Classification (Professional Edition)

Metashape 2.4 adds the ability to train a classifier on a user-labeled point cloud using a Gradient Boosting model .

  • Training attributes: Color, NIR band, Intensity, Return number, Normal, Confidence .

  • Workflow: Tools > Classify Points > Train new classifier; specify a description, the training point cloud, and target classes .

  • Run Classification option allows immediate classification after training; multiple clouds can be queued (classified one after another) .

  • Saved classifiers appear under Use Existing Classifier tab for reuse across projects .

  • HTML reports can be saved when using multiple classifiers (source and target point classes must match) .

Texture: Natural Blending Optimized

Performance optimization of the Natural blending algorithm for texture construction, improving efficiency in producing photorealistic 3D models .