Reconstructing Small Object - All Sides

[Paul King]

Hi, I've been using the free version of Zephyr and mainly use it for small artefacts and coins.
However, I can't get my head around how to include the underside of an object.
Would I be able to achieve this with the 'Lite' version and if so, how?

Here's my latest artefact, but as you can see, it only has the upper surface and thin sides in the reconstruction.

Thank you in advance.

Paul

[cam3d]

Hi Paul King

Coins are very difficult to capture as they are, in their most abstract form, two discs with a thin edge bridging between them. 3DF Zephyr works by matching features between images, and if the features only occupy a small area of the photographs then there is very little to 'bind' them together into one cohesive cluster of cameras from which to reconstruct from.

You can get creative with how you position the coins/subjects on your turntable (vertical stand capture as opposed to laying flat) - Run the images through Zephyr, progressively adding photographs where required. Masking out the non-subject elements of your images is important to eliminate features which could pull the camera orientation the wrong way.

Here's a tutorial on turntable capture if you haven't seen it already: Link

You can send problematic datasets to support@3dflow.net and we can give more targeted feedback on how to improve your image acquisition.

[Paul King]

Hi cam3d,

Thank you for your answer.
I know coins are difficult but thicker items such as the object above are what I'm directing my question at.

So, would I be able to create a 3D model showing all sides?
I've seen this done by lying the object on its side (after photographing the sides and top) and then photographing the base on the turntable. Would the 'Lite' version be able to do this or can it only be done using control points as in the professional versions?

Thanks again
Paul

[cam3d]

Paul King Yes you can use the light version without control points to create full captures including the underside. If you follow the tutorial here, mask out everything but the subject and ensure you have enough overlapping photographs between subject flips and assuming the surface has enough features for the SFM camera orientation phase to successfully compute then it's basically the same as normal processing.

Here is an example of a 50 photo full reconstruction: https://skfb.ly/6APZv - It's an ideal subject as there are lots of features for camera matching so your mileage may vary depending on what you're looking to capture.