A Successful Failure: Converting Google Earth Footage to a Printable 3D Model

I was not sure if I should write this as a How-To guide or as a collection of lessons learned. Having thought about it, I have decided to go the lessons learned route. This means the read will be a sloppy mess and, perhaps, not immediately useful. Hopefully, this is still a worthwhile endeavour.

The reason I decided to not do a How-To guide is because, ultimately, I consider this project a ‘successful failure’. Therefore, I do not have the confidence to write something as definitive as a How-To guide. Before I go into why, though, I think I should explain what the project is.

Nicola Ellis asked me if I could create a 3D print of Cave del Predil in Italy. I said something along the lines of ‘I should be able to’ and got to work. The goal was to use the print as part of an artwork she would exhibit at the ‘Peripheral Memories’ exhibition (April 2021) in Italy's Friuli Venezia Giulia region. I will not go into too much detail on the exhibition here because it is out-of-scope for this blog post. So, I have included a couple of extra links below if you would like to know more about it.

• Nicola Ellis' exhibition profile page
• Udine Today article

I said above, the print is part of an artwork. One of the things Nicola did to the print was coat it with mined zinc powder from the area. You can see the result in figures 1 and 2 below.

One of the things Nicola and I spoke about was the role of the 3D print. My memory is a little fuzzy on this but my notes (I made at the time) suggest there was talk about 3D printing a mould instead of a solid structure. The mould would then be used to make a cast of the Cave del Predil using the zinc powder situated there. So, I started looking into TPU filaments to print a flexible mould. Unfortunately, I had to scrap this idea in the end, because of how the 3D model was made.

I had a look on the World Wide Web to see if I could download a model of Cave del Predil (fig. 3) and I found one at CG Trader; At least, I think I did. Neither Nicola nor I could work if this was Cave del Predil. So, we decided to keep this as a back-up if we could not improve on it. This is where the biggest breakthrough came.

I found the links below, which take 2D images and convert them to 3D. The breakthrough came when we discovered we could use screen-capture recordings of Cave del Predil via Google Earth.

• Instructables Guide (for creating 3D models from images taken from Google Earth)
• Autodesk ReCap (program used in Instructables guide to generate the 3D model)

Long story short, I could not get Autodesk Recap to work on my machine. My final hunch was hovering around my graphics card being the thing it did not like. So, I started to look for alternatives like it. On a personal note, I have come to hate ReCap and will not recommend it. With that said, if it works for you, it works for you. Moving on…

Having scouted around, I found my ReCap replacement: Agisoft's Metashape. You can download Metashape here.

After discovering Metashape, I found this article in Agisoft's Helpdesk Portal on how to reconstruct a 3D model using images. I used this, in combination with the previous Instructables Guide and developed the following workflow:

1. Create a screen recording (fig. 4) of the Cave del Predil area, using Google Earth.
2. Create still-frames of the recording using FFmpeg.
3. Import the still-frames into Metashape (fig. 5).
4. Build a mesh from the still-frames (fig. 6).

To convert the screen recording into still-frames, I ran the following Bash script.

bash # cd into the directory with the screen recording in it. ffmpeg -i predil-video-google-earth-001.mov -f image2 -vcodec copy \ -bsf h264_mp4toannexb "%d.h264"

Apart from the ReCap problems, the project was coming along without too much hassle. Having said that, it is here were things started becoming more difficult.

When I tried to print any model made with Metashape, Ultimaker's Cura – and FreeCAD to be fair – would open them as a mesh and not a solid shape. This lead to my 3D printer printing a ‘phantom shape’. The nozzle would move as if it was actually printing something but nothing came out of it. I think the main issue was the model did not have a height/thickness. So, the printer would move the nozzle to the point where the mesh was but, because it did not have a thickness to it, it did not print anything. There was nothing to print, in other words. I will admit, though, I could be wrong about this, but it is the best idea I have at the time of writing. And, for the sake of completeness, I will note I found a guide on how to create a solid from a mesh in FreeCAD, but it did not solve the problem. This was made even more weird when I could get the printer to print the model's support structure – whilst the actual mode did not print.

In an attempt to fix the ‘thickness problem’, I decided to import the mesh into Blender. From here, I managed to apply a level of thickness to the mesh (fig. 8), but I created a new concern in the process. A lot of debris, for lack of a better word, was lurking over the now thickened mesh. Because of this, I had to remove the excess polygons (fig. 9) manually, within Blender. Otherwise, Cura would not accept the model.

I completed a few test prints using black Polylactic acid (PLA) after I removed the ‘debris’, and this new approach made a difference. It actually printed, but it still relied on the support structures, generated by Cura. The support structures were generated automatically, so it was not an issue in my eyes. With that said, the aim of printing a mould was completely out of the question now. After discussing this with Nicola, we decided to try future prints with a transparent plastic, instead of the black PLA I had used up to this point. The thinking at the time was it would minimise the amount of colour-bleed through the zinc powder. I guess now is a good time to confirm the plan had changed to coating the printed model in the zinc powder. The end result turned out much better than I was expecting – as you can see in the image (fig's 1 & 2) above.

The story does not end here, though. Several more problems still persisted. The biggest one being the change to the transparent ‘glycol-modified’ Polyethylene Terephthalate (PETG). I do not know if I had a bad roll, my printer does not like transparent PETG or (most likely) I missed something. Regardless, the PETG kept warping and not sticking to the printer bed. I tried different supports (E.G. rafts), temperatures (bed and nozzle) and layer thicknesses. The results did improve, but I never resolved the issue completely. I tried printing in PLA again and the result were much better – albeit in black. I, also, ran out of the PLA which meant I was kinda stuck with the PETG – budget restraints and all that.

The next concern I needed to solve was how to increase the print’s rigidity. I never truly solved this one, either. To be fair, I did actually increase it, but I never got it to at a level I was happy with. My hunch is I increased the scale of the model in Cura to a point which causes a similar effect to pixelating a 2D rasterised image, such as .jpg and .png files. I tried various scaling factors and could not find the threshold for where this rigidity issue started to kick in. Each print, which was scaled up, had this problem, though.

For whatever reason, Metashape produced relatively tiny models. Nicola wanted the print to take up as much of the printer bed as possible and, to do that, I had to scale the models up by 3000%. To put things into perspective, the printer bed is approximately 22 × 22 cm. Even at 3000%, the print did not come close to reaching the end of the bed. When I imported the Metashape meshes into Blender, I needed to scale them by a factor of eighty before they were usable. Going from Metashape to Blender and then Cura produced more rigid prints – with a smaller percentage values needed in Cura – than going from Metashape straight to Cura.

Unfortunately, I do not have much documentation to share with you, regarding the rigidity issues. I did not think to document the project’s progress which has left me with very little to show you, visually speaking. On top of that, I no longer have access to the failed models and prototypes. So, all I can do is try to explain the problems as best I can.

Each up-scaled print had a sort of spongy feel to them. The prints felt like their walls were connected by layers of elastic bands. They would squish inwards but spring back, retaining their shape in the process. None of the models buckled or broke, but that is mostly because I did not want to break them. I had, also, ran out of PLA at this point. So, I could not prove it was the up-scaling or the PETG causing of the issue – or the combination of the two. I have, since then, printed other models with a different roll of PETG and not had the squishy feel, so I am inclined to think it was the roll.

As stated previously, there was an intention to print a mould and create a cast of it using the zinc powder from Cave del Predil. I thought this was doable when I managed to give the Metashape mesh some thickness (in Blender). I was, also, very optimistic at one point. It quickly died, though, after I realised fairly quickly the model was too messy – for lack of a better word – to coax it into a form useful for casting. The biggest hurdle I had to overcome here, before adding ‘walls’ to the model, was removing the loose polygons/faces (fig. 8). If you tried to print the model with them part of the file/model, Cura could not slice it or produce the G-Code necessary for the printer to use. Unfortunately, cleaning the model up – after giving it a thickness – was a manual, slow and tedious process. So, I did not have enough time to work out how to extend the outer edges of the model and turn it into a (box-shaped) mould (fig. 11). I did know how to get Cura to generate the support structures underneath the model/mesh, though. This allowed Nicola to coat the print in the zinc powder as an alternative solution. As stated previous, this was not the original desire but this approach did generate the desired effect Nicola was looking for (fig’s 1 & 2).

You might have noticed the final result (fig. 1) is squared-off, while most of the other images show a more organic perimeter. This ties into originally wanting to create a mould. Whilst experimenting, using Blender's Boolean Modifier features, I could square-off the edges. Nicola decided to go with this design decision because it allowed the models to use more of the print bed. Thus, I could make the prints bigger (fig. 11).

This last point is not really a problem. It is more of an observation, which I found odd. After developing the system to produce a printed model from the video footage, I began to increase the fidelity of the video output. This meant I started recording the Google Earth footage of Cave del Predil for longer and at higher resolutions. This, in-turn, increased the resolution of the screenshots fed into Metashape. I, also, increased the settings in Metashape to produce more detailed meshes. Having done that, I found the models did not improve. The video – footage recorded at 720p produced prints equal to the footage recorded at 2K. For the purpose this project, it was deemed acceptable but I still found it curious. I sometimes ponder if the lack of improvement is a natural by-product of this process, or it was something I did wrong. If there was more time and money in the project’s budget, I would have liked to investigate this more. Alas…

As I stated near the beginning of this blog post (or novella?), I view this project as a ‘successful failure’ – from my point-of-view. From Nicola’s point-of-view, the artwork has the effect and aesthetic she was aiming for. So, I cannot say it is a ‘successful failure’ in that respect. Having got this far, though, I hope you can see why I shied away from writing a How-To guide. My hope is you take the lessons I learnt from this project, apply and expand upon them without having to find out the hard way, like me.

To end this blog post, I will say I have made some renderings (with Blender) of the digital model (fig. 13). You can view them in the 3D Models section here.