Photometric Stereo Technique for PBR Materials - Using Shadow to Capture Surface Detail
ฝัง
- เผยแพร่เมื่อ 31 พ.ค. 2024
- #photometric #scanning #tutorial #materials #photography #capture #photogrammetry
In this video I present and share my experience with the photometric stereo technique, which is one of a few basic reality capture techniques used in PBR material and atlas texture creation. While it struggles with large and medium sized shapes, it is probably the best technique to capture micro detail. Its based on a few years of my research, so I really hope you are gonna enjoy it.
If you like it and want to see even more content like this one, please subscribe to my channel, share it and leave the thumbs up.
Big thanks who did it already as it really helps me to move forward.
Be safe and till the next one
Cheers!
Grzegorz
In details I present:
0:00 Intro
5:23 Introduction to Photometric Stereo
8:42 Camera Mount System
12:39 Light vs Shadow Quality
13:41 Props for Testing
17:41 Horizontal Shadow Direction
19:39 Vertical Shadow Angle
22:04 Slim vs. Point Light - Light Shape
25:13 Slim LED vs Point Flash power comparison
29:47 Light Pollution
32:13 Light Bounces and Inverse Square Law
38:20 The Background
42:06 Close Distance Light Bounces
43:28 Cross Surface Gradient
45:31 Scale and Color References
47:07 Chrome Ball
48:49 Color based shape misinterpretation
49:49 Cross Polarisation
52:19 Capturing Transluency
54:10 Workflow - Software
57:52 Summary
Your level of details in process documentation is almost at labolatory level. Please do not stop! I've just spent tonight 3hrs watching your videos. Time to get some sleep and process the knowledge.
Thanks, 3 hours in one go sounds impressive :D. I am really glad to know you found them interesting and useful :D. Cheers!
@@GrzegorzBaranArt i mean to ask yesterday but to tired so forgot. Have you been testing focus stacking technique for macro products?
@@krysc4d Yes, its a lot of additional work and complicates the process even if automated focus rail was used. I belive it all can be automated but I am not a big fan of stacking for photogrammetry captures. Especially that stacking significantly multiplies the number of images captured, therefore affects the camera life span significantly if used with the flash. Camera shutter usage can be reduced by use of electrical shutter mode but in this case needs LED source of light to be used. But it is all totally doable and we can scan this way even with microscope objectives.. its just a lot of work and it gets tougher the smaller subject we are trying to capture as amount of images increases due to shallower depth of view. I might record a video on this suject at some point.. but dont plan it soon as its too niche I guess :)
You make it look so easy , you are a monster , great job and thanks for your time !! the video ( and your channel ) is worthy of the greatest technical tutorial !!🤯🤯
Thank you, veeeeery appreciated
It's just a brain explosion. So many nuances, so many answers. It's really like I bought some kind of guide.) Just a bomb!!!
Thank you, very appreciated
Thank you very much!
You are welcome :)
thank you for sharing this video
You are welcome
Amazing info here. Thank you!
Thanks :)
You're a legend amongst all the other workflow tutorial creators. I am working on a video now and I can't even stomach the amount of effort you put into this one. Bravo.
Cheers Chris, yeah, thats true. I also didnt realise how much work some folks put into their videos until I did a few by myself :). For example, a few sec long shots with coloured balls for this one took me 2 days just to animate and render :). Whats even more time consuming is the research itself.. sometimes it takes me months of trials and errors and thousands of £ just to verify, prove and put a single sentence together :).. but this is how we learn and makes knowledge share even more valuable. Good luck with yours videos
This was amazing, thank you!
You are welcome
I was literally just preparing to try to do some work like this soon, and struggling to find resources - this is a huge help. You're doing God's work.
Glad to hear that Thomas, and good luck with your attempt :)
Very useful video! Thanks!
You are welcome :)
Thanks for your detailed video, appreciate the work you've done so much
You are welcome, hope it was fun to watch
Thanks!
Thank you! Sorry for so late response, but missed that one somehow and just going through comments :)
Ale super materiał, dzięki za tyle wiedzy!
Bardzo milo mi to slyszec :), Pozdr!
Really awesome video. Very detailed and clearly explained, especially the animated diagrams to explain complex parts. I'm sure this video must have been a crazy amount of work to put together so big thumbs up for that and thanks for sharing 👍👍👍
Thank you Barny, veeeeeeery appreciated
the best resourse i ever find, the BIG THANKS to You, greate detail explenation.
Thank you veeery much, I am really happy to hear that
Very informative video as usual! I did some photometric stereo test a few months ago but never created a good setup for it. Good video to help with that.
Thank you for sharing your knowledge and experimentation!
My pleasure, the cool thing is that most of the gear can be shared with other techniques, so unless you dont build a custom fixed rig, its actually quite affordable. I would definitelly suggest to invest into a good light source and honestly.. the TTL flash is really fun and straight forward to use and solves a lot of issues :). Especially that nowadays, Details Capture or ArtEngine have chrome ball based automated light angle detection. Unfortunately ArtEngine supports just up to 16 images .. at least the recent version. Details Capture 256 to compare :).
Amazing video.
THank you, appreciated
Just ordered extension cable to use ar400 as light source. Impressive 1h video. Great stuff as always Grzegorz
:D
Hello Grzegorz Baran! Thank you for the very nice video! I saw it very interestingly! That's great! Have a nice day!
You are welcome and have a great day too :)
I can't believe so much good info is packed into this vid, and appreciate you take the time to explain and demonstrate everything, even if it means making a 1 hour+ video. I'm sure it was a lot of work to edit. I have been trying to learn photogrammetry and I can see so many errors I made, because I didn't know this stuff before. I'm looking forward to doing some better captures.
Thanks, very appreciate. We learn all the time so its normal. The key to excel in something is to never give up, thats it. Btw. this exact video is about photometric stereo technique which is an alternative to photogrammetry, its not photogrammetry. This technique tho can be combined with photogrammetry to extract the surface data on the level photogrammetry struggles with :). Good luck with your new captures :). Cheers!
This is Gold content.
Thanks Raf :), appreciated
Unbelievable effort
Such a kind of tutorial should be a must for everyone who works in CG field
Great thanks
You are welcome :)
Finally! New video🤩
Yeah, it has been a while since this one was really tough to make.. but finally its done so can start with another one :D
Fun stuff man.
Cheers!
This is insane! Another excellent tutorial. Thanks a lot man! Despite I probably stick to the "normal" or "easy" photogrammetry-workflow with editing the PBR-maps by hand and guessing how they should look ^^ But it's extremely fascinating to hear you talking about and explaining it!
Thanks man, whatever works for you is always fine. I shared it since its good to know and understand other options out there. There is another one I plan to cover .. an AI based reconstruction based on a single image, which is also a promissing option if well approached. Anyway, each technique is just another tool :). Cheers!
@@GrzegorzBaranArt Awesome! That would be great to see your AI-approach on that topic! Keep up the great work man!
Wreszcie 🤣 super tutorial
Dziekowac :), ciesze sie ze sie podoba
@@GrzegorzBaranArt duzo skanowales tą techniką? mój setup jest bardzo podobny, ale zastanawiam się, czy nie lepiej byłoby zbudować zmotoryzowany "scan rig" jak Dave Riganelli. Po zeskanowaniu kilkuset liści na serio odechciewa się ciągle przestawiać to samo światło dookoła, uważając żeby czegoś nie dotknąć albo przerwócić :)
@@dainjah Mysle ze jesli skanujesz setki lisci i wszystkie sa mniej wiecej tej samej wielkosci, do tego skanujesz zawsze w tej samej lokalizacji, automated rig, z dodatkowa opcja cross-polaryzacji to jest totalny no-brainer i dlugo bym sie nad tym nie zastanawial. Natomiast glowna wada takiego rigu jest brak jego mobilnosci i skalowalnosci. Zeby zeskanowac cos pstrykam przewaznie okolo 20-30 zdjec. W przypadku rigu, oznaczaloby to 20-30 zrodel oswietlenia. Do tego lepsze rezultaty osiagam czasem z podluznym swiatlem a czasem z punktowym (flash).
Wiec rig na pewno ulatwi prace i pozwoli na jej automatyzacje, natomiast jesli planujesz jezdzic w okreslone lokacje i bedziesz dealowal sie z obiektami w roznej skali, taskanie takiego rigu ze soba, transport samolotem itd. zaczynaja byc problematyczne na tyle ze w praktyce wybiera sie inna technike. Oczywiscie mozna zaprojektowac bardziej mobilny rig z tasmami ledowymi pre-cross polaryzowanymi, mozna ograniczyc sie tez jedynie do 4 czy 8 katow itd. Co sie liczy na koncu, to jest rezultat. Nikogo nie obchodzi jak go osiagnales jesli jest bardzo dobry. Wiec tak naprawde to Ty wiesz najlepiej czy rig to jest wybor dla Ciebie czy nie. Nic tez nie stoi na przeszkodzie zeby miec dwie opcje w zanadrzu. Zgadzam sie ze reczne przestawianie swiatel jest ryzykowne i meczace na dluzsza mete, natomiast to swego rodzaju kompromis. Uwazam ze jesli masz na to budzet, to dobrze zaprojektowany i wykonany statyczny zautomatyzowany rig na pewno zapewni Ci lepsze rezultaty przy mniejszym nakladzie pracy i czasu.
@@GrzegorzBaranArt Próbuję zrobić zautomatyzowany rig z 1 światłem (przy pomocy arduino), ale to jest już inny świat, a ze mnie słaby programista i elektronik ;) chyba lepiej jest zainwestować w 16 żarówek z filtrem polaryzacyjnym i mieć spokojne nerwy 🤣
Thanks
Omg, thanks Thomas
thanks
No probs, anytime :)
thisis soprofessional tips for photography
Cheers
This is a gold mine, thank you sharing it! I am struggling to find software linux-compatible. I saw on a forum thread that Meshroom's dev were interested in supporting photometric stereo, let's pray for it to happen! I wish I could help it
Glad it helped!
Great Tutorial Grzegorz!
Just wondering how to proceed with the Data.
Are you planning a second part on the photometric stero reconstruction part?
Anyways great video! Thanks a lot.
Thank you
@@GrzegorzBaranArt Thanks a lot, I will have look on your artstation!
Which tools would you reccomend for seemles textures like you mentioned at around 57:23 ? Trying to figure out what to use for bark textures.
It usually depends on the subject, texture size and budget. I recommend Unity ArtEngine for this purpose and this is something I use on a daily basis. Unfortunately while it was the best tool available for seam removal, it was discontinued and cannot be purchased anymore. The next solution I would recommend is probably the manual seam removal in Substance Painter - I covered it in some of my previous videos and in details in my first book about photogrammetry. I plan to record the entire video to cover seam removal subject if I ever find some time and enough motivation. There are also other solutions like PixPlant, Sampler, Designer, Affinity Photo etc. and some are better in some areas while some in others. Even half automated offset based manual seam removal in Photoshop or Affinity is also an option and seems to be the only solution when you deal with extreme big texture resolutions, Painter, Designer or ArtEngine wouldnt even handle. Hope that answers your question.
From the listen software of photometric stereo, which one is your preferred solution for processing the multiangle images?
Thanks for the amazing tutorials!
You are welcome. For photomettric stereo reconstruciton I would recommend Details Capture. I use it on a daily basis and imo its the best option from these I am aware of at least. Cheers!
For the light directionality, why not use a light source with a snoot? It is used to get 'shafts' of light like god rays. That is the kind of thing you are looking for right?
Not sure how would it help unless you meant it as the way to reduce the amount of environmental light bouncing (pollution). Bear in mind that the subject needs to be kept in cone of light anyway.
Super informative video and great thanks mate! Would be even better with another tutorial introducing the software workflow😂
Thank you
wow, awesome amount of information - on the "straightness" of shadows, would a fresnell lens do something about straightening it, it also seems to have a more linear falloff.. I stripped some off a TV while looking for polarisation filters (they are difficult to take off tvs as they are pasted on like a sticker, but you get 2 fresnell lens films)
Thanks, honestly, I have no clue, I would give it a try for sure as long it doesnt affect the beam purity. I believe that all depends on the scale. If you scan 15x15cm area and the distance between the light source and the subject is safe, you shouldnt get too much angle distortion. For 50x50cm area coverage, the angle spread might be a big issue to solve as the considerable working distance would be way too long. As said, definitelly worth to try, as every single improvement really matters here.
Great vid as always. What is this color spectrometer that you use here? 46:54 Is it good choice?
Thanks, I would say that its the subject of my another incoming video about color calibration as its a very wide, quite complex but interesting subject :). Good spot btw :), was wondering if anyone is going to notice
Hello howdy, very good video thanks for your time to explain, I wanted to know what features your pc has.
Cheers, sure, my PC isnt anything special. Its I7-7700K and 64GB of RAM with GF1080Ti on board
@@GrzegorzBaranArt Hello, thank you for your reply.
I wanted to know if I can send you a project that I made is a photogrammetry test, to see if you can give a little retrolimanetacion, I am interested in my pc is an i7 - 8750h with 15 of RAM and a gtx 1060 6 gb graphics card, and well doing the job with 6 million polygons as practically spend 5 hours in the whole process, I know you handle dense models of polygons and well I think in relation to the pc I have is to increase the RAM to 32 as it is a laptop on which I work. but I think I can get a little more out of it, I remain attentive and thank you in advance for your training on the TH-cam channel.
Very nice tutorial. Have some questions:
1. How do you extract specular/roughness maps (polarized vs unpolarized images ?) , AO ?
2. I think ArtEngine can do 16 images, Adobe Sampler 8, but can this be done in Zephyr?
1: By substracting data between corss and non crosspolarised captures. But for generic surfaces I simply generate roughenss using technique I shared in my videos already. I use albedo, ambient occlusion/BlueNormal channel, or use the node which generates roughness/gloss based on color ArtEngine offers.
AO can be generated from the height map or from the normal map. Blue channel of normal contains kind of AO map.
2. Yes, ArtEngine photometric stereo reconstruciton is limited to 16 images only. Details Capture to 256 limited also by the amount of available RAM, both Adobe products, Sampler and Designer are limited to 8 images only. As far as I know, Zephyr has no photometric stereo module but during computation generates noise on a micro level from the albedo. Its a fake noise really and I would say.. more a single image based one. Photometric Stereo is more accurate and reliable regarding the captured data accuracy. I would say that 3D Zephyr is made purely for photogrammetry based reconstruction which is a different technique to photometric stereo as is based on totally different concept.
@@GrzegorzBaranArt thanks for the detailed answer. All the best. Looking forward to new content
You may consider black velvet across your surface rather than spray paint. It's a good light absorber and can be cleaned easily with a lint roller or tape.
Thanks, unfortunately I did. I have found that fabric gets dirty way too easily and is tougher to clean to solid surfaces. Also fabric has too much details as a surface itself. Even cardboard was too detailed. Photometric stereo is so detail oriented that I was regenerating all this fabric noise beneath the actual subject ;). So I have found that the cleaner and more polished the surface, the better, the only problem was to make it matt so it doesnt reflect environment :). As I said in the video, I dont cross polarise side shots so every black surface is going to be visible to some degree. The point is to make it clean so its easy to cut - talking about vegetation and overall litter captures. For tiling materials, fabric as a light absorber should be fine
Great video, would you mind explaining how to apply photometric stereo using Ds and Sa?
Thanks. Initially I was planning to record another video with detailed comparison of most common photometric stereo apps, but since they are under constant development and change so rapdly, I wasnt able to catch with them up and gave up. I also didnt notice much interest in this matter on my channel :). But I might reconsider it at some point if there are more folks interested. Cheers!
was just the best video on photometric stereo! how was your result using phone flash as a light source? is it complete garbage? i am thinking about it for very small leaves!
Thanks. Light is the light and depends on spectrum it covers, its intensity and clarity, but also on environmental light pollution, distance from the subject and everything I mentioned in this video.
I plan to make a video about light sources but need to save about £3000 for a special devide I need to use to measure spectrum detail so it is goint to take some time.
Anyway, imo the best tool for such captures is the wireless camera flash light with TTL feature to auto-adjust light intensity, as well as some decent, real camera equipped with lens, mounted on solid platform which prevents camera movement and micro-shake. Of course any light can do the job if used properly. Sure, you can use your mobile and get some results and it all depends on what do you really need and what is enough. To be honest, sometimes you dont even need photometric stereo, just some 'single image to PBR' processing tool which does the guess on values - or you dont even need a PBR data except the albedo -for example this is how I made this flower for the glass vase scene here: www.artstation.com/artwork/WBO8qQ
@@GrzegorzBaranArt Hey thank you so much for such a detailed explanation! I tried with image to material in Sampler, but imo photometric stereo prevents the blobby cloudy details made during image to material. I have not yet started a job in the industry, so a bit short on supply haha . Therefore I am considering buying a cheaper solution, I don't need very high fidelty, but it should be passable! Tried with a ring light with very soft shadows, did not turn out great! So I will have to look for some more I guess!
Have you ever given an IR flock sheet a try for the BG? Or try painting something with the notorious "Blacker than Black" paints?
The way both materials basically work is how they are meant to absorb a serious percentage of the light hitting it.
This may be more expensive in compare to a series of other options but the results should be a bit more of a guarantee.
Nah, I haven't. I was considering but they were too expensive and the setup I build was good enough already. I have noticed that any paint fades in time and gather some dust when used.. therefore its purely black just for a short period of time :). I am sure that blacker then black paints should work better and make captrues easier.
So I was considering to purchase one and even had it in a cart already but.. it was £50 for 100ml while the cheap one I used was £7 for 400ml.. and I had some other things in my 'must purchase' queue so I deprioritised it :)
I was wondering when your next post was gonna be!
Cheers! Had the same problem as this one took me really way too long.. but I hope it benefited from the this time and effort taken :)
Hey, I have a syrp turntable motor and I was wondering if it were possible to rotate a texture surface instead of the light, then realign the photographs in post-process?
Unfortunately I dont understand your question. With the turntable you dont move the light or the camera, but the subject you are trying to capture. Camera, same as the source of light stays in fixed position in this case.
@@GrzegorzBaranArt Sorry to respond so late, I'm just coming back to this video. Maybe I should have watched this video in its entirety but did you go over the practice I postured? What you said in your response is exactly what I want to do, so maybe there's a part of the video I should watch.
Hi. This technique looks ideal for recording apotropaic marks in buildings I'm surveying. I use Agisoft Metashape for photogrammetry, but haven't come across the software for PST. Can you recommend? I use a Mac computer. Thanks
I covered a bit about the software in this video already. I considered a separate video where I was planning to cover more details but I dropped an idea do far due to lack of interest. Regarding photometric stereo apps, each differs and offers something other dont. I prefer Details Capture because it has no image limit for processing. ArtEngine is also really good but it has limited amount of images it can process to 16 only - at least for now. Both of these offers chrome sphere based angle autodetection. Substance Designer and Sampler are limited just to 8 input images and expect fixed angle therefore. They dont have chrome sphere based angle autodetection and therefore are way behind the competition. Of course Designer offers some awesome tweaking tools as well as option for easy atlas composition. Substance Sampler .. at this stage with these tools it has seem to be a waste of money and wouldnt bother with it.
@@GrzegorzBaranArt Thanks.
Hey, great video! The best source I've found about shadow capturing & photogrammetry :)
I'm thinking about buying the godox Ar400. What are your impressions about using the Ar400 as key light for shadow capturing? Would you recommend to buy the ad200 too or do you think the Ar400 would be good to start with?
Thanks -
Justus
This exact video is about photometric stereo technique, not photogrammetry. Just saying. For photogrammetry I would strongly recommend AR400. For photometric stereo AD200. AD 200 has TTL option which compensates uneven distance to the subject when shooting around. It is also smaller as a source of light while AR400 works more like a aerial light. To be honest, any light should do for the beginning, .. some use even mobile or simple torch. Of course it affects the final quality. I recommend flash mostly because it can overwrite the environmental light so I dont have to work until it gets dark to capture something :). Hope that helps, Cheers!
Yes I know :) Just looked both of your videos 😊
So probably I'm going to buy both. I'm a 3D Artist and would love to capture my own materials :) haha spend the last days to figure out what light setup the guys from megascans are using but the elinchrom elb 500 setup is right now a bit to pricy and I don't see a great advantage (except +100ws and TTL in comparison to the Ar400) (when I understood it right that you can use each output at 250ws when using both ^^)
Thanks a lot :)
Can you please let me know what software is available for photometrics reconstruction - I'm not familar with the Sa & Ds icons in your video. I use Metashape for photogrammetry and have a Mac. Thanks
Sure, I covered it in this video already in part about the software :)
Have you tried light modifiers (especially on flash or high-powered light sources) like grid or wrap the end of flash with thick black paper/cardboard as an extension tube of sorts?)
Why would I? I used polarisation filters to get rid of glare but glare is not a big issue with enough angle coverage and the light source placed under 45 degree angle.
@@GrzegorzBaranArt i meant for a more channeled / direct light light source for your photometric work or working with shadows. I have not heard of this fascinating technique before but i was wondering if it would work well for this technique :) .. maybe not… I don’t know.. just throwing it out there. This or cone light modifier was something i used a lot in portrait photography to channel light ..minimizing spill.
@@vrguytokyo I think that many depends on the exact application and its algorithm used for data processing, so if the app wasnt designed for something, even better illumination might give oposite results. It was the issue I had with slim light as none of apps I used were desinged with this type of light source in mind. But I would consider a great idea for whatever works and gives you better results as et the end, this is the result what matters
Hi Grzegorz,
A good friend of my sended me the link to this video as I'm trying to deep dive into texture/atlases scaning.
Also been working for over a year in scanning 3D assets with pretty decent results, but found hard to get the results I wanted when dealing with scaning 2D surfaces. I just wanted to thank you for all the info you shared here, and also if you don't mind, ask you about your Ring Flash polarizing filter holder and sheet as I have the same PixaproRiko400 and didn't find a proper one to solve the task.
Can you tell us if it's available to purchase or is any kind of self-made asset? Thanks in advance!
Hey, sure. I suggest to watch the other video where I covered it better: th-cam.com/video/REA3XNgUMJg/w-d-xo.html
Its a magnet mount SCANSPACE filter and is available for purchase from their store. Hope that helps :). Good luck with scanning
@@GrzegorzBaranArt thanks mate! I just ordered mine :D. Very kind of you to share!
Sure, I’ll take a look at your video. Keep up the good work…appreciate the effort!
is it possible to combine photometric stereo scans with a photogrammetry scan of the same object in order to increase the detail level of photogrammetry scan. If so, how would you do it?
Yes, it is possible and some folks does it on a daily basis. Its not so simple tho, as it combines two different techniques, therefore needs dedicated rig with doubled functionality and dedicated software to merge 3D and 2D data.
Could you try BLACK 3.0 ? In both photogrammetry and Photometric Stereo. It should help a lot if what they claim is true.
I don't have access to exact BLACK 3.0 to try it out but to be honest any black surface with high absorbtion level will do. Some will do slightly bettter to bother, but the way you use the light and set your camera is more important in long run. Of course better absorption level always helps and makes the job easier.
Will your new book come out in 2023, will it be an update of the previous one or will it still be a new product with a new price?
Hey, I am still working on it as I don't want to release it until its not completely ready and covers every single aspect I believe should be there. If nothing changes I plan to release it in 2023. It is going to be a new product with new price but kind of based on the first one. Since some folks asked me about color calibration which is part of the incoming book, I plan to release just this chapter seaparately earlier - probbaly end of this or beginning of next month as part of incoming video about color calibration.
which software do u use in 2024 for photometric stereo?
Still Details Capture
Do you think the new DJI Mini 3 pro will work well for capturing surfaces?
It depends to what you compare it to. Any standard camera outcompetes any drone and will give you better results. Also as I understand Mini3 has a sensor with just 12Mpx of effective pixels.. to compare Mavic2Pro has 20. Since it is lighter it is going to struggle with wind a bit more to heavier drones, so its a bit more weather dependent. Wind has especially big impact for close shots.
Honestly I would consider this drone.. similar to Mavic2Pro as a flying mobile. It will do the job where camera cant and provide you coverage camera might struggle with. It has worse parameters to Mavic2Pro but they are still much better to old Minis and old Air drones. I am sure that if properly used with enough coverage it should do the job well enough.
Especially that drones might be super useful with new AI to single image AI based algorithms incoming as they allow to capture larger area easily
@@GrzegorzBaranArt Thank you for your input. The main reason Im thinking about the mini 3 is its light weight making it almost regulation free in the EU where I am and the Mavic 2 will be heavily restricted. I have a DSLR that gives me great results. Maybe I should try the DSLR workflow on my cellphone as a way to see what I can expect from a Mini 3. I have learned a lot from your videos, quality stuff that is much appreciated!
@@PcKaffe Any camera is better to no camera and any drone is better to no drone. Of course there are exceptions and old Mavic Air was so bad that was basically useless. Mini 3 is much better compared to its predecessor and should do the job. What I mean is that I would capture areas you can reach with the standard camera. And I would capture those you cant reach easily with the drone.
Its a good idea to test quality with your mobile.
Does anyone know of some good 3D reference; to have a detailed digital 3D model and then actual 3D object that one could scan?
That would be cheap to buy as a package, or some worldwide spread comercial product where I can buy the 3D object in a grocery/store and get the 3D model online?
Like coca-cola bottle or something even better - more detailed and complex, any ideas?
Hey! What camera would you recommend for photometric stereo capture?
The one which has the highest image resolution (more Mpx) as in this case texture size unlike in photogrammetry, is directly dependent on the actual image size. Second thing is the lens. Actual brand doesnt matter really but if I have money and start clean, I would probably go with Sony A7R4
@@GrzegorzBaranArt Expensive as hell! I will experiment with some entry level DSLR camera to see what results will i get. Shame that the lenses are more expensive than the body sometimes!
@@Igor_Zdrowowicz oh, but you didnt ask me what affordable camera would I recommend :D. I also can't afford Sony A7R4 but I had access to it and would recommend it for this purpose. Regarding lenses, they dont get old as quickly as camera does. If you capture a lot, you need to change camera every few years because it has very limited life span. Sensor gets more and more hot pixels .. not to mention that each camera has limited shutter life span. Lens stays longer, there are many folks who use even very old (20, 30 or even 40 years old) lenses with proper adapter. I wouldnt recommend DSLR anymore as basically all camera producers has discontinued DSLRs. For example old DSLR lenses for Canon dont work anymore with new RF mirrorless Canon mount without EF-RF adapter. 3rd party lenses are cheapper and often worth to consider as an option. THis is also the reason why Canon doesnt allow for production of any 3rd party lenses for their new mount.. which is probably another main reason they lose so many customers recently. But maybe it will give them a lesson. In practice 3rd party lens can be even 10 times cheapper to its brand counterpart.
@@GrzegorzBaranArt Oh i see this field is not free from the greed politics! Ofc my bad not mentioning that I'm looking for an entry level stuff 😅 I'm reading a lot opinions about Canon G12 being good for texture hunting but i guess rarely anybody not already knowledgeable in the camera quirks and features means photostereoscopy and capturing perfect crisp normals with it.I guess it would be cool to have some entry level camera, good enough for large scale textures, as well as those sweet saturated plant leaves and their normals.
Greetings Master. What apps do you use for reconstruction?
It depends. I recommend Details Capture and ArtEngine. Details capture has no image limit, has chrome sphere based angle autodetection and gives the best results but lacks of additional supportive tools. ArtEngine offers similar reconstruction options and good reconstruction quality but is limited just to 16 images. From the other side ArtENgine offers some additional useful tools. You can also use Substance Sampler or Substance Designer, but both are limited just to maximum 8 images and dont have chrome ball based angle autodetection (fixed angles only).
So for reconstruction I would recommend Details Capture and this is my primary tool for this purpose. Next I usully use ArtEngine to remove seams, generate roughness by subtracting cross polarised image from parallel polarised one etc. At the end I use Photoshop/Affinity Photo to tweak values if needed
@@GrzegorzBaranArt the only ArtEngine i find is a discontinued project for unity. Is it that one?
@@X3msnake Yeah, this is the one I mentioned. Its just an option though, there are many other tools/solutions out there. For example you can consider PixPlant for automatic seam removal, or Substance Designer, or Substance Sampler, or remove then in Photoshop, Painter, Affinity manually etc. From reconstruction point of view, 'Details Capture' is the one which provides the best results from these available on the market (regardign these I tested).. at least at this time. It might change in a future
@@GrzegorzBaranArt in my case I am mote interested in the rconstruction alone to use for high detail capture on subjects like coins where Photogrametry and 3D scanning have a hard time or are slow. Did u try any opensource solutions like Fiji (imageJ) photometric stereo tools
@@X3msnake As I see its an open source which is nice.. unfortunately I haven't tested it. Is this good?
18:35 umm, have you heard of fresnel lenses? They do just that... Or a parabolic reflector.
if you're trying to get shadow detail, why would you use the camera flash? 24:40
Because this technique is based on observing shadow when cast from different directions. The flash light allows me to illuminate the subject with the light strong enough to ovewrwrite any environmental light but also to shorten cameras exposure time. I covered details in this video. I can use a LED light instead but LED light needs longer exposure, therefore isnt strong enough to overwrite any environmental light and needs to be used in total darkness. The camera flash I use in here isnt connected to the camera but is free to move around.
Last but not least, any photometric stereo app expect to be fed with this type of data for reconstruction. Hope that makes sense :) but I would suggest to watch the beginning of the video where I covered it in detals already
This looks like how GreyScaleGorilla clay materials are made
Would Metashape solve this kind of dataset for reconstruction?
Edit: Zapomniałem, że mogę po polsku pytać ;)
I am not sure what dataset you are askigng about? For photometric stereo we capture series of images from exactly same spot. In this case position of the camera doesnt change. From photogrammetry point of view it would be just a single image illuminated from different directions, therefore would be useless for photogrammetry reconstruction. Technically we need at least 3 images captured from different positions (under different angle) to estimate accurate position of the subject in 3 dimentional space.
Metashape, 3DF Zephyr or Reality Capture are photogrammetry tools. This video isnt about 'photogrammetry' but about 'photometric stereo'
@@GrzegorzBaranArt That's exactly what I was asking - actually between my question and your answer I tried it with some fabric and it didn't work in Metashape but worked great in Sampler ;) Thank you for the great work by the way - it is tough to find know-how about PBR scanning and photogrammetry.
@@notorioushax I was planning to make another video to show how to proceed with these captures, but wasnt sure anyone is interested. Instead I just dropped the 'workflow' section at the end of this video (around 55th minute). There are many apps with 'Photometric Stereo' functionality and Metashape isn't one of these. You can use Details Capture, ArtEngine, Substance Sampler and Substance Designer instead. Each has its pros and cons. Cheers!
fleshlight or flashlight?
:D
34:09 anxiety
:D yeah, I guess I should add a lens too to hold the fabric stronger ;)
Cyborg
Cheers!