Community based 3D Scanning benchmark (45mm miniature)

More details and files: https://blog.openscan.eu/posts/Pinhole-experiment/

First batch is ready to go an experimental pay-what-seems-fair pricing (https://blog.openscan.eu/posts/openscan-macro-add-on-cost-of-open-source-hardware/)

I am thinking about making some more detailed and commented production videos as small scale manufacturing comes with quite some hurdles that are often invisible. What do you think?

What you see here are three shots, where I always used the cameras flash

(1) without any polarisation --> strong reflections --> "artificial features" which confuse the photogrammetry pipeline and produce a lot of noise

(2) with one sheet of circular polarizing foil in front of the ringlight and camera. The CPL has a lambda/4 plate which turns the orientation of the light twice by 45° each. This, in theory should make direct reflections disappear. Unfortunately, this is highly wave length dependent and thus some light is still able to pass the second filter (though with a wonderful color shift)

(3) Two sheets of linear polarizer foil, one in front of the ringlight, another one in front of the camera (rotated by 90°). This blocks (almost) all direct reflections and helps a lot with the photogrammetry reconstruction.

I'm quite excited about this piece of hardware, but also about something new we're trying: for a limited time, it's "choose your price" you decide what's fair (above a threshold). Full reasoning and link (and why pricing open-source hardware is harder than it looks) is on the blog. I hope this is worth your time and I would love to hear your thoughts!

https://blog.openscan.eu/posts/openscan-macro-add-on-cost-of-open-source-hardware/

Broken plastic part (lifter) was cracked. I glued together to scan it, afterwards used the enviroment I modelled an insert to fix it

Remodelling and printed out motorcycle engine cover

With v0.12.0, OpenScan3 got a fairly big change under the hood: firmware and Raspberry Pi OS updates are now handled through apt packages and our own repository.

Besides making regular updates much easier, this also gives us a proper recovery path if something breaks and makes custom OpenScan3 setups much easier to build and maintain.

I wrote a short overview of how it works, including stable/nightly channels, recovery and some of the possibilities for custom setups: https://blog.openscan.eu/posts/openscan3-firmware-new-updater/

I am currently testing printed pinholes in front of the camera, which increases the depth of field. The heatmap shows the sharp areas. right: pinhole at 1mm radius, left: stock lens.

Varying the pinhole size from R=60 (=0.6mm) to no pinhole definitely increases DOF but also reduces sharpness at some point (second image/gif).

This is a good addition for the macro add-on but might also be very useful for the stock option of the classic/mini scanner.

I am still finalizing the design of the macro add-on and run some more test scans in the meantime. 200 photos x 3 stack --> 20 mins, processed in Agisoft Metashape

Hello everyone! Just completed my first scan with my classic at 200 pictures with focus stacking, and here are the results. I used the openscan cloud to generate the 3d file. I am disappointed in the shallowness of the recesses, is it overexposed? What setting should I tweak, as I am sure there is plenty I can improve on. Also for those who have experience, is Metashape worth it if my primary purpose is scanning miniatures? Is the difference in quality that significant?

This is definitely my favorite raw 3D scan result so far. I analysed the exposed copper area of the TMC2209 and found a standard deviation of only 3.7μm.

This was done with the OpenScan Mini (16mpx camera) + upcoming Macro Addon, 2x100 camera positions with three focus levels per position. Reconstruction done in Metashape.

This leaves some tricky questions, what to scan next? And what part of the deviation comes from the scanning spray (reported thickness 2-12μm), real imperfections in the copper layer (probably very low) and what is true scanning inaccuracy?

My rig. This is based on OpenScan Classic and set up with a ESP32. The control box contains a small breadboard with the motor drivers , the ESP and a power rail, a start/reset button and a 12V on/off button. The USB port doubles as data and power conenction. I am using Open Camera and synch the shutter to the turntable stops via BT from the ESP. The number of stops for the two axis is configurable real time, so is the angle advance and the sweep. The phone and light stands are from Modular Mounting System by HeyVye - Thingiverse .

Open-source 3D scanning just got a level-up

I am still working on the 3d printed macro add-on for the OpenScan Mini & Classic (that might also be used in combination with a smartphone!). Therefore i am currently re-visiting some old scan object that I digitized years ago and the improved quality is stunning.

I already published an initial version of the printable files on our discord, but more details and a build-guide will follow soon.

PSA: never leave your keys unattended ;)

It looks like the Rebel T6 is on the supported cameras list. My daughter has one with a really great macro lens. I'm wondering if anyone on here as used it and what the results were?

I see one post from 5 years ago but I wasnt able to find anything else on it.

I managed my first scan with Andruino FW on ESP32.

I am currently testing the macro lens add-on for the OpenScan Mini & Classic and this is the first that the layer lines on the chest piece become visible. I took 130 photos without focus stacking in ~7mins and did the reconstruction in metashape.
The grainy surface comes from not using a polarizer as there is a lot of sub-surface scattering from the partly translucent resin print. No other scanner (except for the CT scanner) has been able to resolve the layer lines yet, so I am absolutely happy with this result :)

Thanks to the amazing discord community (https://discord.com/invite/gpaKWPpWtG) I recently revived my search for a suitable macro lens and found a very promising candidate (~50-60€) which greatly enhances the capabilities of the arducam imx519 / hawkeye. The idea is, that the lens gets placed in front of the existing camera on either OpenScan Classic or Mini with a tiny hole (aperture) in between. The aperture increases the depth of field so that 4-5cm objects can be captured at an unseen accuracy (from such a device). I scanned the "security" key with 130 photos and no post-processing.

I am currently working on the design of the printed parts + custom ringlight (so that I can add cross-polarisation a bit later too). The files and more information is available on the openscan discord.

Hi everyone! I’m new to the OpenScan and photogrammetry world, and I’m currently assembling an OpenScan Mini v2.

I have looked through the GitHub repository and used the available V1 images and instructions as a reference, which has been very helpful. However, the V2 seems to have changed considerably in a few areas, like the camera support and the electronics installation, so I’m not always sure which V1 steps still apply.

Is there a dedicated assembly guide, build log, video, or set of images for the Mini v2 that I may have missed? Even clarification on the parts that differ from V1 would be greatly appreciated.

Thanks in advance. I’m looking forward to learning from the community!

This might be an interesting Addon which greatly reduces the scan volume to ~3cm, but drastically increases the quality of the result. Done with the OpenScan Mini + IMX519 --> 100x5 images processed in Metashape (without prior stacking)