My first gaussian splat test on the POP 4, nothing mind blowing but I think it came out pretty well considering the lighting conditions weren’t ideal as I did the scan at night in low light. For reference the camera is about 10cm (4in) across, so it’s fairly small and compact. The Pop 4 did a good job capturing the smaller details even though it’s not really designed for scanning small objects, it’s more for medium sized objects from my understanding, but still impressive. This feature definitely sets it apart from the Inspire 2, which is my favorite 3D scanner, but that may change soon as I test the Pop 4 more.
Another test. This time I scanned my wife’s turtle jewelry case measuring about 7 inches in length (about 18cm) using the laser grid mode for a more detailed scan, but I did do a separate color scan for gaussian splatting.
The laser grid works really fast (especially in GPU mode) and I was able to scan the turtle top to bottom in about 1 minute, this would have been unheard of with other scanners in this price range just a couple of years ago and it even outperforms my Inspire 2’s IR parallel line laser mode in terms of scanning speed by a noticeable amount due to having significantly more laser lines which means less back and forth sweeps due to a higher rate of point cloud collection.
I’m definitely having fun with the Pop 4, I will be working on a review video that I hope to have release sometime in the next 6 weeks or so, but in the meantime I will post my test scans here.
The textures look amazing. You’re underestimating the results when you say they’re nothing impressive. Of course, they’re inferior to photogrammetry, but it requires better preparation with lighting and is clearly more time-consuming.
Thanks for the response and I’m glad you think the scans look great. Maybe I could have worded it better. I do think the scans I posted are very good, I was just trying to emphasize the poor lighting conditions in my setup and the fact that I know the Pop 4 can do even better in terms of gaussian splats if lighting conditions are optimal. The Pop 4 really is a great scanner that can work quite well in different lighting though
It don’t requires more preparation with lighting compared to Photogrammetry and is much faster in processing especially when using POP 4 , because the half work is already done while scanning the object volume .
I never had so much fun with Photogrammetry in the last 26 years as I have now with 3DGS
Most amazing things are not just the textures , but also semi and semi transparent materials , Photogrammetry lossing to it on so many levels .
The magic of perfect 3DGS is not really lighting but capturing all angles possible and not leaving dead spots , the software itself will correct the lighting automatic .
However for best colors a good LED lighting is required and I am not talking about the brightness
Perhaps there’s a language barrier. I’m saying that classical photogrammetry requires good lighting to produce good results. Then it would outperform this technology.
I do agree with you , yes it does, but the results of classic 3DGS already are beating photogrammetry not only on the visual part but also the geometry is much better. Finally something moved forward since 30 years …
I thought I would share this scan of my old shoe. I was quite impressed with the detail it managed to capture, including all of the imperfections and wear the shoe has experienced. The scan went buttery smooth.
I’ve used The Pop 4 long enough to get a feel for it, and I want to say in many ways it feels like a turbo charged Inspire 2 with the addition of being able to create gaussian splats. What I mean by this is that scanning can be done in less than half the time with the Pop 4 vs the Inspire 2. Both scanners are designed for medium sized objects, and both have a laser mode which means real life scan quality and detail appear to be relatively similar at normal viewing distances, but I would give the edge to the Pop 4 when you zoom in a bit as it has the more robust laser scan mode which does edge out the Inspire 2, but both scanners deliver great scans.
I’m definitely seeing the value of the Pop 4 as I use it, and I understand why its priced higher than the Inspire 2. It’s not so much that the 3D scans themselves are vastly superior to the Inspire 2, but it’s the speed at which you can scan objects and the addition of creating gaussian splats that really sets it apart. I think I can officially say the Pop 4 is now my go to 3D scanner, though I can see myself still using the Inspire 2 where ultra portability is a factor (like sticking it in my pocket) or if I don’t want to disturb people around me as the IR laser is invisible whereas the Pop 4 laser is quite bright.
Here is a 3DGS of a Nerf Toy Gun. It came out looking great and details hold up reasonably well even up close (though it does get blurry if you zoom in too much). The scan has a few minor issues that I can easily fix by making sure I’m a little more thorough when I scan, and I didn’t have uniform lighting while scanning so there are bright spots, though I did do this as a single scan. Speaking of which I noticed after a firmware update today that I no longer can create a 3DGS from a mesh that was created from merged point clouds. I can only make 3DGS from a mesh created by a single scan. All of my past 3DGS were created from merged scans so hopefully this is just a bug.
I wanted to challenge myself and scan something a bit more complex. I chose to scan my DJI gimbal as it has pretty complex geometry relative to the other things I’ve scanned. I think it did a great job considering this was just a single scan, I did not merge multiple scans to create this 3DGS.
@Generation_YFi Hi Chris, since you own and use both the POP 4 and INSPIRE 2, could you compare their real PC and memory requirements?
With the same medium-sized object and a similar final point distance:
Does the POP 4 require noticeably more RAM during fusion?
Does its 30-cross-line mode provide the required coverage with fewer captured frames than the INSPIRE 2?
Are POP 4 projects generally larger despite the shorter scanning time?
How much RAM does your computer have, and what is the largest POP 4 scan you have successfully fused?
I mainly scan smooth composite parts around 50–70 cm long over USB on a Windows laptop with 24 GB RAM. With my INSPIRE 2, fusion has already reached about 96% memory usage, so the POP 4’s real-world memory requirement is an important part of my purchasing decision.
Hello. Im sorry for the late reply. I’ve been super busy. I will try to answer your questions, but I will need to do a test scan and do a direct comparison of the Inspire 2 and Pop 4’s memory usage to say definitively.
However if I had to make am educated guess, the Pop 4 will consume more system memory and vram in laser mode. I say this because the Pop 4 is capturing more point data per second than the Inspire 2 while also running at around double the framerate. This is why the Pop 4 can scan an object in half the time or less of the Inspire 2. Real world scan quality is similar between the two, but the Pop 4 has the edge because of its dense point clouds that it can create in quick fashion.
Does its 30-cross-line mode provide the required coverage with fewer captured frames than the INSPIRE 2? This is tricky. Yes it captures more data per frame, so it does technically require less frames to get the appropriate coverage. However it also runs at a much framerate so you end up with a lot of scanned frames regardless.
Are POP 4 projects generally larger despite the shorter scanning time? I’ll need to check, but I would assume Pop 4 projects are larger, but probably not drastically larger since the post processing phase is good at getting rid of redundant and overlapping point data.
How much RAM does your computer have, and what is the largest POP 4 scan you have successfully fused? My computer has 64GB of DDR5 system memory and 16GB of GDDR7 Vram. As far as largest Pop 4 scan I did, it would probably be a mini fridge. I actually scanned a larger object with my Inspire 2 which was a 4 foot tall (1.3 meter) dresser. The Pop 4 would be able to scan it no problem too.
Overall the Pop 4 is basically a faster Inspire 2, so unless you want faster scanning the end result and capability is similar between the two scanners in real world use. The main draw of the Pop 4 is the gaussian splat (3DGS) feature. It makes much higher quality color scans that could be used in videogames, video content, or preservation. The difference in quality in that regard is miles better than the Inspire 2.
Thank you, Chris. To clarify my situation, I am considering replacing my INSPIRE 2 with a POP 4, rather than adding another scanner. My main use is reverse engineering in Rhino 8: I use the scan mesh as a geometric reference for reconstructing composite parts. I do not use 3DGS, and color quality is not important to me.
My laptop is a Lenovo LOQ 15IAX9, model 83GS00MMHV, with:
Revo Scan rates the GPU as Excellent for the POP 4, but the CPU only as Good. More importantly, an INSPIRE 2 scan of a smooth composite part approximately 50–70 cm long has already pushed total system-memory usage to about 96% during fusion.
For my Rhino workflow, the important question is therefore not only whether the POP 4 captures the object faster, but whether I can actually benefit from that speed with only 24 GB RAM. If its higher frame rate produces larger raw projects or causes fusion to exceed my available memory, the overall workflow may not be faster or even practical on my laptop.
When you have time to perform the direct comparison, could you please scan the same medium-sized object with both scanners, aiming for comparable coverage and using the same or a similar final fused point distance, and record:
the number of accepted/captured frames;
the raw Revo Scan project-folder size before fusion;
peak system RAM usage during fusion;
fusion time;
the fused point-cloud size or point count, if easily available.
The scanning paths do not need to be perfectly identical. Two normal, practical scans intended to produce comparable Rhino reference geometry would be more useful than a laboratory-style test.
Based on your real-world experience, do you think replacing the INSPIRE 2 with the POP 4 makes sense specifically for Rhino-based geometric reconstruction on this laptop? Or would the 24 GB RAM limit prevent me from gaining much of the POP 4’s speed advantage?
It’s been a little while since I’ve logged in here, I initially wanted to post 2 to 3 Pop 4 scans a month, but I had an unplanned move out of my apartment that I had to deal with and a major health incident in the family that needed a lot of attention. Things have settled down a bit so I can now get back to using the Pop 4 scanner again. While unboxing my stuff at my new place I came across a Jurassic World T-Rex figure that I thought would be perfect to get back into the swing of things.
I’m still super impressed with the Pop 4’s ability to do gaussian splats, the texture detail of the T-Rex figure really is something to behold. For a scanner meant for medium sized objects it still is impressive to get this level of detail. I can’t wait for Revopoint to launch a 3D scanner meant for scanning small objects that can also do gaussian splats, that would be awesome to get even higher levels of detail while up close. As an aside I also noticed a new feature to add reference images to help the gaussian splat feature add in additional detail, I haven’t tried that yet but I’m looking forward to seeing how that works.
Your system is the minimum requirement you can use and you may run out of memory if you set your fusing point too high , the minimum is 32GB of RAM , the recommended RTX is 4060 , RTX 3050 is minimum requirement.
CPU is on the edge of minimum too , so not fast system but you can do the work to some degrees, but it will not provide you the maximum speed and performance especially if you scan with laser modes. For Full Field it is more than enough.