Are there any plans to improve software tracking?

Is that a question ? Suggestion ? I think you should ask Revopint , they programming it after all .

Hi All,

Thank you for all the discussion and reply.

I have sent this post to our developer team to discuss. We will reply when it is convenient.

Best Regards
Cassie

Thanks for the info. I can try to specifically formulate ideas considering tracking - not just “Make it better!”, but “You may want to try this and this” like not allowing software to instantly “rotate” the scanner between neghbouring frames, testing it’s speed between previous and next frames, fixing possible drifts in the turntable mode and alike in case your programmer if he think he might have a use of them. Not 100% sure I can make useful algorithms, but I can try it.

The Revo Scan software will keep acquiring data in the hopes that tracking will be re-acquired and synchronization will take place. This is standard behavior.

When it doesn’t happen, Stop (maybe followed by Undo) is all I’ve needed to get rid of it, although the unsynchronized (red) stuff remains on the screen until I get a proper refresh.

In my opinion this should be a priority for Revopoint instead of continuing to develop new hardware each half of the year, such a business plan of overpromising and underdelivering could’ve worked (and did work) in a competition free space but with major players, such as Einscan, entering the game they should finally focus on the software.
I have used products of the competition and the only thing Revopoint is better than them is scan resolution, but that’s quickly overshadowed by the ease of use of the others (and hello, but we’re talking about consumer/hobbist grade stuff - that means its no.1 priority should be ease of use!).
Even Creality quickly catched up and although the scanning resolution is meh (~0.5mm) it’s actually possible to hand scan with it, not to mention the software that let’s you edit your scans INSIDE the project.
I don’t want Revopoint out of the game as they did start the recent popularization of 3D scanning technologies in hobby fields but without major software and algorithm improvements it just is a possibility.

Here’s some of my proposed software related things that should be changed or added:
-Fixed marker mode; Currently it’s using the silliest implementation possible. Firstly instead of detecting the markers on 3D data the algorithm first transforms it into 2D image, where only then it begins searching for circles, yes perfect circles. When scanning you usually point the scanner at an angle and thus the markers turn into elipses in the 2D view; this is the reason why you don’t always get all the visible markers appearing as “red dots”. The solution to this would be to detect the markers in 3D space, this would also fix the second issue which is the often tracking loss. You see Revoscan computes the marker locations in XYZ space but doesn’t compute their normals, because of that sometimes marker pairs can get matched to a wrong pair resulting in misalignment. By tracking the markers in 3D you’d also be computing their normals thus solving this problem (If XYZ locations are close but the normals are not you won’t get a match).
-Texture/hybrid scanning modes; I’m not talking about texturing the scan but using the texture of our object/surroundings for 3D tracking. All Revopoint scanners have a RGB sensor but it’s ever used for coloring the pointclouds while it could be used for extra tracking stability. Of course you’d need to use it in a well lit environment (well maybe besides the Mini), as POPs don’t have a flash lamp, but that would solve most of the crazy scan drifts from happening.
-Using host’s compute power; If I’m correct all Revopoint devices have a chip inside that basically does all the bundle adjustment and stuff, and then streams it over to the host. But what if the host also did some of that processing? It doesn’t all need to be done on that poor chip, for example a PC could exclusively have the above scanning modes extra (although a modern phone should be able to handle them with ease too, just look at their AR performance, perhaps even make use of that AR API when inside the mobile kit, hmm?). I think it’s already being done with marker mode as it needs a direct USB connection, then why not have an option to completely skip the chip and use only the host’s resources? Other scanners do just that and they’re doing perfectly fine, also with USB 2.0 speeds.
-Merging Revoscan and Studio together; Other scanning suites already have everything in one package, not only does this speeds up the processing of scans but also saves time as for example you already see that two scans don’t fit together and can outright redo the wrong one.

With just a few of these the “Revopoint experience” could soar high, so please consider them our lovely Revo engineers if you’re reading this! :>

3 Likes

This was attempted (and Revopoint personnel spoke about it here in the Forum), but it didn’t work, so it has been shelved; at least, until a new approach is developed.