3D model, Perfection in a single object

I want to know if there is an option to focus only one object in the 3D model, so that it comes out with better quality and HD, is there any way to do it?

manual? Automatic?

You could upload just the pictures in which the object appears and ask DroneDeploy to process that small part with higher resolution. But I think without a Business or Enterprise Plan this may not be possible. The other option is to process the photos yourself using something like Photoscan which offers a free 30-day Pro trial with full functionality. When used with its highest setting, Photoscan typically generates 3D models with 4X the resolution of the default DroneDeploy model. But it will run for hours if you do not have a good computer with a high-end graphics card like the Nvidia GTX 1080 Ti (which is on sale now that the new RTX cards have come out). This approach does have drawbacks: it has a learning curve, is more complicated than just using DroneDeploy and a license costs as much as a 3-year Pro Plan subscription.


1 Like

I would recommend using the House & Roof, Precision 3D Model app. Use the tools of the app to try to reign in a point of focus.

1 Like

I’d fly manual with sequential overlapping oblique images. I did that on this one and it came out fine. You can see the image pattern by investigating the map.

1 Like

Thank you very much, I hope it works for me.!

Is there any tutorial on how this application is used? I already have it but I do not have the knowledge to use it.

A couple things to keep in mind. I agree with Gary. Manually obtained oblique images. Try to have very little or no motion while image is being captured. While you may not “see” the difference in the images when you look at the jpeg, the processing algorithm sees things the naked eye can not see and will matched those points up when it processes. A better camera (lower GSD) or lower altitude/distance from point of focus will help. The final and probably most important concept to understand you are seeing a model (data/colorized triangles ot some geometric shape) you are not looking at RGB imagery. You are seeing colorized meshed between data points. So it will never be as crisp as a true RGB image. I know this is a DD forum. As others have mentioned, there are other processing options out there which give you a lot more control over how your data gets processed (such as Pix4D) local processing using custom templates. If you understand what a model is, how it is made and the limitations you have then combining imagery obtained from obliques and nadir and as many of us who have been doing it for a while, experiment with lighting (times of day and weather) as well as varying altitudes. Remember there are passionate “modelers” out there who only use hand held cameras and even iphones who obtain incredible models. This field is really a combination of science (photogrammetry) and art/experience. Experiment when you can resort to proven when you need for $ projects. Follow the above guide to get a lower GSD/higher resolution/overlap especially in the areas you need to focus on.


I’d like to add to that. When you think in the facts of GPS accuracy it is better to take further apart shots in One direction and then take them again back to where you came from. If you take images to close to each other that can cause a ribboning effect. Instance for the face of a building I will take a picture on a 50% overlap, no more, until I reach the end and then reverse my path and do the same thing.
One more thing that I have figured out recently is that in structure mode you need to have your perimeter at the correct distance from your face of subject. You typically want the center of the face of the building in the center of the picture. I know that a lot of people may think that the entire building should be in frame.

1 Like