Agisoft Metashape Professional 2

Modern workflows often fuse photogrammetry with LiDAR. Metashape Professional 2 natively supports importing LiDAR scans (LAS/LAZ) and aligning them with photogrammetric point clouds. This allows you to texture LiDAR geometry with photorealistic color from your RGB images, yielding the best of both worlds: accurate millimeter range data from LiDAR and color texture from photogrammetry.

For enterprise users, the Python API in Professional 2 has been expanded with new classes for network processing, automated GCP (Ground Control Point) detection, and custom AI model integration. This makes Metashape a pluggable component in automated drone-to-3D pipelines.

Agisoft has integrated a lightweight neural network for semantic segmentation. Version 2 can automatically classify points into categories: ground, vegetation, buildings, and water. This is a game-changer for GIS professionals who previously spent hours manually cleaning point clouds. Additionally, the mesh reconstruction algorithm now produces smoother surfaces on organic shapes while preserving hard edges for man-made structures—all without additional post-processing.

Objective: To automatically classify high-density point cloud data into "Vegetation" and "Ground" classes using a multi-stage algorithm (Progressive Morphological Filter combined with Point Density analysis), and subsequently generate a clean Digital Terrain Model (DTM) without vegetation artifacts. agisoft metashape professional 2

Target Audience: Surveyors, Forestry Managers, and GIS Professionals using Metashape Professional 2.

Key Improvements over Standard Workflow:

The most profound technical divergence in Metashape 2.0 is the native integration of machine learning models directly into the photogrammetric pipeline. Traditionally, dense point cloud classification was a rudimentary process based on geometric attributes (e.g., height from ground), often resulting in noise and misclassification in complex urban environments. Modern workflows often fuse photogrammetry with LiDAR

Metashape 2.0 introduces a Convolutional Neural Network (CNN)-based classification engine. This system does not merely analyze geometric coordinates; it interprets the visual context of the imagery during the classification phase. By training on vast datasets, the software can now semantically segment point clouds into distinct classes—such as vegetation, buildings, roads, and vehicles—with a significantly higher F1 score than previous heuristic methods.

This semantic understanding extends to the Masking Workflow. In industrial scanning or artifact preservation, backgrounds constitute a significant source of noise. The 2.0 update utilizes deep learning for automatic background masking, effectively separating the foreground subject from the environment in the alignment phase. This reduces reconstruction artifacts and minimizes the "floating noise" often associated with complex scanning setups, streamlining the generation of water-tight meshes for engineering applications.

The jump from version 1.x to version 2.0 brings numerous headline features that address long-standing user requests. For enterprise users, the Python API in Professional

The most significant architectural change in version 2.0 is the shift from Python 2 to Python 3. This is not just a version bump—it unlocks modern libraries (NumPy, SciPy, OpenCV) directly within Metashape’s console.

How to use this: You can now run complex machine learning pre-processing scripts without leaving the application. For example, you can write a script to auto-mask vegetation using a pretrained model before alignment.

Pro Tip: Use the built-in Metashape.Tasks module to batch-process 100+ chunks. Create a template script that imports os, Metashape, and time, then loops through a folder of new datasets every night.

Upon launching Agisoft Metashape Professional 2 for the first time, users will notice a flatter, more modern UI. The historic “Workflow” dropdown is now a logical left-hand panel guiding you through steps: 1. Add Photos → 2. Align → 3. Build Dense Cloud → 4. Build Mesh → 5. Build Texture. However, advanced users can still open the “Expert Settings” panel, which reveals granular controls for keypoint limits, depth filtering modes, and adaptive camera model estimation.