Skip to Main Content

Photogrammetry Demo in Agisoft Metashape

How to use Agisoft Metashape to turn photos into a 3D model

Agisoft Metashape Photogrammetry Video Tutorial

Step by Step Guide

Step 1: Photogrammetry is the process in which you develop numerous photos into one 3D model. Make sure that when you are taking your photos, you keep all setting the same (ie. zoom) as the software is trying to piece together all the images.

Step 2: Launch the 'Agisoft Metashape' Program on your computer.

Step 3: The first thing you should remember to do is save your project. Select File > Save As. You want to make sure to do this in the beginning, so as to not lose any progress you make along the way, especially in the case of your computer crashing. 

Step 4: To add photos Click 'Workflow' in the top left of the menu bar and 'Add Photos'. Select all of the photos you wish to incorporate into the model to create one big chunk. Hit 'Save' one more time. 

Step 5: To match all of the photos correctly go back into Workflow > Align Photos. This allows Agisoft Metashape to align points of the photos and extrapolate where the cameras where to approximate what the 3D Models will look like. In the Align Photos box, select High Accuracy with the Generic Preselection and click 'OK'. You can choose different accuracy settings to create either more or less detailed models.

Step 6: Let the Program Run to process the photos, this may take some time so be patient. Once it's complete, hit "Save". 

Step 7: The model may appear to be on its side, you can change your viewpoint and left-click and drag your mouse to rotate the image and take a look around, again this does not actually change the actual grid of the model. In order to pan the model, right-click and drag to center your view. Rotate and pan the model to whichever orientation you want to see the object more accurately. If you wish to actually move the object, select the object tools icon  and select 'Rotate Object". This will allow you to see inside the space more accurately.

Step 8: You may notice extra data in the background, these are produced by background materials in your photos. If you want to view the approximations of various locations of where different photos were taken, select 'Model' > 'Show/Hide Items' > 'Show Cameras'. You do not need to see the cameras while deleting data so once you are done, make sure to turn them off to focus on your model. 

Step 9: To remove all of that background data, use the 'Free Form Selection tool' located in the Selection  toolbar in the top left corner. Click and drag your mouse around any and all of the unwanted pieces and hit the 'delete' key on your keyboard when you have selected data you want to remove. Additionally, continue use your navigation tool  to rotate around the model and see what background data is still needing to be removed. 

Step 10: To zoom in and out of the model, use the scroll wheel on your mouse. 

Step 11: A few key things about viewing on this software, you can view the object in 'Perspective' or 'Orthographic', Model > View Mode > Perspective/Orthographic to specifically look at the differences in distance.

Step 12: Additionally, in the Preferences section, you are able to select your GPU, or what type of stereoscopic display you desire to use (we have it set to Anaglyph); this can be found in the General section of the box. Anaglyph allows you to see the model in red-blue glasses, so in 'view mode' you can see the object in red-blue 3D

Step 13: So far, you have created a sparse cloud from your photos. In order to produce a dense cloud, go back to Workflow > Build Dense Cloud. Once that box pops up, select Ultra High Quality and hit "OK'. You can again choose different quality settings at this step to produce more or less detailed models. This processing will take a while so be patient again while it runs. 

Step 14: Once the process is complete , select 'Dense Cloud' in the top toolbar (9 dots in a square). 

Step 15: Cleaning the noise around the model. There are a couple ways to do this, first go back to 'Free Form Selection' to discard the data that's a bit further from the model. Again, just drag around whatever you want to get rid of and delete it. Hit 'Save' once you get the majority of the outer areas, then zoom in to get into the more complex areas. Make sure to get all of the outliers you can in this step before moving on to transforming the simulated surface into a real surface.

Step 16: You can also decide how much of the image is truly essential to your model. For things that aren't crucial to your image but contains a lot of noise, you can go in with Model > 'Rectangular Selection tool' and cut out the excess and hit 'Save'.

Step 17: With an orthographic view, the perspective is more straight on and therefore, much easier to see where there is more noise and what it is that you're trying to get rid of.

Step 18: Select 'Workflow' > 'Build Mesh'. Make sure the options selected are "Dense Cloud", "Arbitrary", and 1,000,000 faces. Depending on how complex your subject is, you can specify larger or smaller face counts.

Step 19: Select  icon to view all three versions of your model to see how much data and depth you've got. Make sure to save all of your progress up until this point.

Step 20: To close up any holes in your model go to Tools and choose 'Mesh' > 'Close Holes'. 

Step 21: The final step for your model is to go to 'Workflow' > 'Build Texture'. In that box go for the "generic" model, the mosaic default, and two 1024 x 1024 textures, then hit "OK' to save and exit. 

Step 22: To see the changes you just made, go back into the  options and select 'Textured'.

Step 23: To save go to File > Export > Export Model. Create a file and save. In the 'Export Model' box the only boxes that should be checked are: Vertex colors, vertex normals, and export texture, and click "OK'. 

Step 24: Right-click on the OBJ file and select Open with > 3D Viewer (this may take some time to load the program). 

If you have any additional questions, please contact