Fun with SFM – part II: The results
So, hello, dear all, I am back from a very busy family weekend – and I would like to entertain you now with our second part to our SFM-tutorial!
Due to my work at University, I have the opportunity to work with Agisoft Photoscan, which is quite the same as Agisoft Metashape. You can find all the information on this software here – I am very used to it, it is very easy to work with, very easy to understand and you hae all the features collected in one software. Of course, there are other softwares, even a lot of freeware for 3D modelling – I used to work with them as well and I think, the next step will be a comparison of the different softwares by modelling the same object. I am just looking for the time to do it, because: My dissertation has to be (okay, should be) finished by December… so, time is really rare in my life right now…
But this post is all about my little Eros/Amor riding his hippocampus. You can find the information on this example here in my first post.
I opened Agisoft Photoscan and loaded all the images I wanted to use meaning all the pictures I have taken. The first step is the aligning of the images. The software searches for the same points on different photos and matches them.

Step 1: Aligning the photos

Step 2: Building the point cloud
The next step will be the point cloud – of all the points that the software could match on different photos (the best thing is that you get the same point at least out of three different directions) there will be a so-called point cloud as a result. As you can see in the second pic here, there are a lot of “noises” around my object – you can barely recognize Amor’s foot an the tail of the hippocampus. So, the next step is to clean this mess up. You just mark all the noises around your object and delete it. Another possibility is to mask your photos before aligning them and starting to build a point cloud. Either step will take its time.

In my first post I mentioned the different directions you have to take the pictures from. Here is a screenshot of all the thumbnails of my pics, just to help you geting an idea.

And again, the front view of my model, here already “meshed”.
So, after cleaning you will get a very nice point cloud showing just your object. I just took the side view of this sarcophagus, so my object remains quite flat with no other sides.

After the last step – the finished model!

In detail and without texture the model consists of a grid that connects all the points with each other and gives the actual shape of the surface.
After building the so-called dense point cloud (you may notice the features of the relief getting clearer), we can start with the last step – to build the actual model. As you can see here, I really tried quite hard with the texture as well. The texture can be build out of all the coloring information you have in your pics. This step is not necessary, if you are just interested in the shape of the surface. Therefore, you will get better results without coloring your model, because your eyes might get betrayed by all the shades and colors.
If you want something nice to show your students or your friends, then go for the texture. 😉
Of course there are further things to do on Agisoft Photoscan/Metashape or with other software – it’s a playground, so feel free to try. And remember, I took these images with my phone camera, so no pressure on that. Of course, your first tries may suck, but hey, it’s a process. I am still learning too and I am still discovering new fascinating features and possibilities.
I hope you enjoyed this little post – stay tuned, keep calm and start 3D modelling. 🙂
All the best,
Astrid
OpenEdition suggests that you cite this post as follows:
Astrid Schmölzer (October 13, 2019). Fun with SFM – part II: The results. Epigrammetry. Retrieved May 18, 2025 from https://doi.org/10.58079/og64