Anytime :) btw. 3DF Zephyr got much better since the time I released this video and is able to generate much denser meshes with better level of micro details. But the workflow is the same.. just saying as re-tested it recently to see the progress it made.
Very very in depth tutorial with explanation of why and what for each step. Too bad that they are not all free tools but I did get a lot out it. Thank you for the effort.
Thanks and you are welcome :). Regarding the tools used.. very often you get what you pay for. In my opinion free solutions are great to start with, but at some point most are not good enough, lacks updates, new features, stability, performance, support, quality etc. Imo the best solution is to purchase perpetual license of the software you are planning to use in long term.. even if it costs a bit more then a subscription .. this way you wont end up without any software when you stop paying. Also what I present is an optional workflow.. I would say that there are key tools you have to pay for sooner or later.. and tools which can be replaced which its free alternative. Its good to know them all as this way you can consciously decide what workflow suits you the best :) Cheers!
@@GrzegorzBaranArt I totally agree with you. Specially in buying perpetual in stead of subscription. I won't be getting any of this at this moment but the work flow can be used differently with other software but with more steps and less quality for a beginner
This is why I make my videos so you can save some time I had to spend.. and this is why it takes me so long to make each as sometimes.. a single simple sentence cost me weeks of actual research :D
Man, that's nice one! I am from Ukraine though, i have evacuated to west part of it, and there so many surfaces and nature objects to scan!! Thank you for that workflow ❤️😀
So I thought it would be fun, right, to try and find the exact wall on Google Street View. Y'know, not really any particular reason. A bit of the geoguessrerer in me. I've spent TWO HOURS walking around South Shields. I'm not even from South Shields. I've never been! I tried to find the landmarks, find the building supplies company that the sign is pointing to (next right), I had narrowed it down to the exact area. But the wall wasn't there! I kept closing Google Maps thinking "you're going mad, please stop", but the urge to find the f'king wall was taking over. Finally I had a lightbulb moment... maybe the wall doesn't exist anymore? I put street view into "back in time" mode, went back a few years, and suddenly, in that curve of road that now stands nothing but empty sky, the wall! The building with the green trim. Those billboards. It was a suddenly breathtaking moment of clarity. But then of sadness, because that beautiful brickwork is sadly no more... But then another moment of joy, when I realised it's been forever immortalised as a PBR texture. So thank you, for letting that wall live on forever more. goo.gl/maps/5oTwhbnPF5TSDKDB7
Its an amazing story :D and I am impressed you managed to find the exact spot :). Ive just checked on google maps and you are right, apparently part of this exact wall with the building was removed recently. To be honest I have realised that its been over a year since the last time I was there in person. Cheers!
Hey, awesome work and awesome tut. You mentioned you have a plane with a special UV, can you explain a little more about it? how you did and what for? Is it just scaling up the plane UVs until it's greater than the 0-1 space?
Hey, thanks a lot. Nah, the plane I made covers additional 10% for each side outside of the standard 1:1 coverage. It allows to remove side connection points coming from the offset which would be super hard when simple 1:1 plane with 1:1 uvmapping is being used. I described all the details of this plane in my ebook ( www.artstation.com/artwork/K0PYG ) and added this plane to it for those who might use it. I think I also covered quick preview for it in one of my videos but cant remember which one.
Love the tutorials Grzegorz. including seagulls :) I'm currently working on a sandstone Priory (church) and would love to to see how you'd approach old building materials and surfaces. I've not been given drone approval yet on the building but may be able to next spring. (Malvern Priory)
Thanks Darran.. especially for watching till the end. Regarding building scanning, it depends on the size of the height and size of the object but it is totally doable. With the drone it would be just very easy and can save a lot of time and pain :) compared. Without the drone .. unless it is not Notre Dame .. I would consider to use similar technique to presented in this video.. just instead of 2m long monopod and cable remote I would use wireless remote and 'Telescopic Monopod Camera Pole'. You can get even a 10m and loner one. Yeah, would b e cool to make a video to cover the process :). Cheers!
@@GrzegorzBaranArt Thank you so much for your response. Please, please do consider making a tutorial, it sincerely does help us non pro's to start the daunting process.. also The Witcher, i bet you have a story or two of that process :)
Ahh yessss, seagulls and drone footage :> What about surfaces that have really big elements? Like brick wall created from 50cm big bricks or that waved metal sheet from video end? Most of time workflow will be the same, but make it tillable would be harder?
Yeah. Its hard to tile larger elements and depending on the usage, at some point larger element should be driven by the geometry itself. So for example when you scan a cliff wall and planning to use it in a game etc. you should build a complex geometry but cover it with a smaller tiling material. To keep surface interesting you can use mask and mix a few materials together. Dunno if that answer your question :).
Hi Gregorz nice video. Great like always. Thanks for share your knowlege i appreciate it so well. At this time i m still working in the Bark Texture Baking for my own personal tree creation. When i finish my work on that i hope share it with you to know your opinion about it, for me is very important. Cheers..!!
you should do aerotriangulation alignment before processing photos, then replace the original photos with those processed photos for reconstruction texturing
Hey, thanks for the tip, but could you please tell me why do you believe I should? This isnt a large aerial capture with inconsistent lighting conditions etc. I usually dont have any problems with image alignment as for materials I get very even image and coverage control points distribution, consistent lighting, consistent angle and distance to the surface. So I cant really see any reason to complicate the existing process.
I was always interrested in how to make accurate roughness map. It looks like behind that is something more than photogrametry workflow. For me it is one big mystery. Of course there is a way by mixing all other materials into one, but it's not accurate and i'm trying to find the right way.
Yeah, roungness is very tricky since it is a kind of PBR hack. I plan to cover it in series of videos. I started them but didnt have time to finish any yet.
Great videos, very informative. Thanks for your time and effort. Btw can you please make a video comparison between Agisoft, 3DF Zephyr and RealityCapture with the same photos ?
Hey, thanks. Yes, I have it planned and even created one, but never released in form of any video :) because I decided that it is worth to wait for 1.7 update of the Metashape which is going to improve medium and high frequency noise reconstruction, and also for an update from 3DF Zephyr which is going to bring huge improvement for med and high frequency reconstruction for photoconsistency pass. I was also planning to include MeshRoom in this comparison :). So since incoming updates are going to change a lot I decided to wait. Currently, Metashape and Reality Capture delivers similar resut.. its hard to judge which one is actually better. 3DF Zephyr fails on med and high frequency details. Metashape is the slowest one tho, 3DZephyr and RC do the job quicker.
@@GrzegorzBaranArt Thanks a lot and good to hear. I have zero knowledge of Metashape unfortunately but was counting on RealityCapture. Hoping to see upcoming video when the updates come.
Honestly fella Ive been watching a few of your videos now and the amount of effort you put in is incredible!!! I really thank you for the insight and inspiration you provide to me for the field of photogrammetry... Thank you very much. I would like to ask a question. How do you manage your files? I think it would make a great video in itself. How to be organised! Thank you again.
Thank you very much :) doing my best. Regarding the file structure, its definitelly a very good idea for a quick video :) but I guess it depends on preferences so thought it might be too obvious. Basically I have 2 main folders. One is called - SCANS and another one TRIPS. When I finish any capture trip I create a NUMBER-TRIPNAME folder (number is a consecutive number) in TRIP and throw everything from this trip there. In this folder I create subfolders with the gear I am copying from name - for example Canon80D or Drone or DJIOsmoPocket or Mobile etc. DjiOsmoPocket and Mobile usually contain documentary files. I take a lot of these as references but also as B-ROlls for my videos or shots I can use in my incoming book. Next when I want to process exact scan I create a folder with its name and consequtive number in SCANS folder into a RAW subfolder. When I process them in photoediting software I drop them into another folder called .. lets say PROCESSED_TIFFS. Next I process scan and store the main file there. When I export scans and textures I store them into EXPORT subfolder. I bake into a BAKE subforlder. I also have subforlder with documentaries I share on my Artstation profile. When the material is ready, I save it into a subfolder with the texture name. So to give you an example: D:/TRIPS/ --/ 101-CoastStonesAndGravel/ -----/Canon80D/ -----/OsmoPocket/ -----/Drone/ -----/Mobile/ --/ 102-DockyardConcreteAndGravel/ -----/Canon80D/ -----/OsmoPocket/ -----/Mobile/ --/ 103-ForestTrip-GrassAndTrees/ -----/Canon80D/ -----/OsmoPocket/ -----/Mobile/ D:/SCANS/ --/ 432-Gravel/ -----/raw/ -----/docs/ -----/tiff16bitsCalibrated/ -----/bakes/ -----/gb_gravel432/ -----/export/ -----/gravel432_scan.xx (photogrammetry_project) --/ 433-Gravel/ -----/raw/ -----/docs/ -----/tiff16bitsCalibrated/ -----/bakes/ -----/gb_gravel433/ -----/export/ -----/gravel432_scan.xx (photogrammetry_project) --/ 434-Rock/ -----/raw/ -----/docs/ -----/tiff16bitsCalibrated/ -----/bakes/ -----/gb_rock434/ -----/export/ -----/rock434_scan.xx (photogrammetry_project) etc. I keep this structure well organise since I backup all these files from time to time as photogrammetry related files takes a lot of space. So I have HDDs which stores scans from 1-133, next 134-230 etc.. and similar to trips 1-30, another HDD has 31-42 another next and so on. Hope that makes sense :) Looks like I dont need to make a video to cover this anymore. Cheers!
@@GrzegorzBaranArt Superstar fella!! Most Kind. Thank you for such a well considered response, that I really didn't expect to receive! So very Grateful... My organisation, in honesty, is my (but not my only) weakness... and I will be correcting this. I have been interested in 3d Design and photogrammetry at a very novice level for about 10 years... And at last I may have a nice project to begin my own portfolio, we will see. Thank you again.
Большое спасибо. Желаю и вам всего наилучшего. Оставайся в безопасности
4 ปีที่แล้ว +1
perfect tut! thanks! I just wish they implemented some easier and more precise simplify/retopo solution directly to Zephyr/RC to avoid the high number of post-softwares we need...
I totally agree. So far you can use simple decimation but it often sucks with more complex meshes. Also to make a texture you need the canvas to be as square as possible to avoid space waste. Also I guess as it looks straight forward for flat surfaces it might be an isue for complex 3D meshes. But I totally agree that simple lowpoly retopology tool with the option to align/snap to the heavy mesh might be very helpful.
I wouldnt use 50mm at all for capture like this one.. maybe35mm would also do the job.. Especially not from such close distance. I would say.. if I shot from a larger distance, and the surface is less complex and I use the tripod for camera stabilisation 50mm or even longer makes sense. But since I shot the surface from really close distance, the surface had a lot of depth and I used a monopod handled in a way you can see on this video I prioritised a shutter speed and camera's beam view over frontal data density. Basically with the 24mm mm more light gets into the camera and the shutter speed needs less time to take an image so the final images is sharper. With the 50mm I would get more blured images and less side coverage. The data captured for this reconstruction and the image coverage is way beyond of what I need for reconstruction anyway, so setting used was totally correct. Bare in mind that camera distortion doesnt matter if you know how to use it. Its just a tool and has its own pros and cons. I would agree that lens set to its edge values brings some additional consistency distortion.. but it is still ok for this type of capture. I aimed to create 4k texture set from it.. THe data I captured is still enough to create 8 or maybe even 16k. Cheers!
Very interesting as always! Is the primary reason for the switch from Metashape to 3DF Zephyr because you can get more fine detail on the mesh? How do you think the two packages compare?
I didnt switch to 3DF Zephyr :), I am just trying to play with different tools to present different workflows.. ther pros and cons. Regarding the comparison.. I am planning to make a video about this and compare Meshlab, Reality Capture, Metashape and 3DF Zephyr. Planning to do this after Metashape will sort issues with depthmaps and release 1.7 but also after 3DFlow guys finish the improvement for the Photoconsistency pass .. hopefully with the next update. So currently when I compare the software I think that Reality Capture gives the best results but is the most expensive and the most impredictable. If I score software from 1 to 10 I would say that RC is 9. Metashape is 8 but incoming updates are going to push it to 9 too. 3DF Zephyr is currently 6. After a new update they should jump to 7... maybe 7.5. And they are also very affordable. Bare in mind that 3DF Zephyr evolves rapidly.. every change is a big change. RC introduces just super small.. slight changes .. so they basically dont move up anymore. Metashape grows faster to RC and they might beat them soon. Regarding the price and sale model.. I would say that Metashape with the incoming updates and CUDA support is going to be the best and the most affordable piece of photogrammetry software on the market :). 3DZephyr is moving really fast but its still signifficantly behind. So if I have to pick between the Metashape and 3DZephyr.. Metashape definitely wins.. but its good to know many tools as it gives you option of consious choice.. and bare in mind that all these tools are in constant change. Something which is decent today.. might become the best option tomorrow :). So you should always pick what suits you the best.. and keep an eye to other options as everything changes very fast. 3DZephyr is a very solid piece of software. Their software is build on a very strong and solid coding base which is hopefully going to pay back with the stability while growin up .. while other software might get stuck at the same time as every new feature might hit the stability as is hard to imlement into a messy code.. just guessing to give the point.. as I said in my video.. there is no 'the only one' or 'the best tool'. All are in constant change.. and you should pick the one which suits you the best. Hope that makes sense, Cheers! :D Dmt.. didnt know if I answered your question.. today I would pick the Metashape.. but would keep an eye on 3DZephyr.. now I did answer :)
@@GrzegorzBaranArt That's got to be the longest and most thorough reply I've ever seen on TH-cam! 😁 You're totally right about always keeping an eye on a range of software so we can always keep improving. Inspirational!
It depends. I have two lenses. For this certain and most of my recent captures I use Canon EF-S 18-135mm Nano USM lens. I used to use an Aspherical Sigma 24-70mm EX which is bigger when compared and gets brighter and sharper images with less distortion. But since I have found that Sigma is a bit to heavy, it focuses a bit slower when compared with Canon and when facing down due to its own weight the gravity untwists the focal length ring way to easily during the capture I decided that darker but lighter and better made canon zoom works better for me Bare in mind that Sigma I mentioned is a full frame lens while the camera I use has cropped sensor. So this 24-70mm is cropped into smaller sensor. Anyway, I was considering to get 24-70mm Canon instead but it seems to be a bit too expensive for me. The Sigma lens due to its size is tricky to be used together with the ring flash light, especially that zooming extends the length of the lens. When used with the AR400 flash light, if not perfectly aligned, extending focal tube can push the polarisation filter from the flash light off :) I didnt cover scanning with the flash light in my videos yet but when I do I need to remember to show the issues large lens can cause. I was thinking to use a prime lens since they are better in every way? But because zoom lens gives me more options during scanning, for example complex surfaces usually need more coverage and higher focal lenth (35-55mm) while simple and flat surfaces can be captured quicker with wider fov (24-35mm). Some surfaces with hard acces might benefit from a telephoto zoom (70-200mm) while shoting on very short distance like the brick wall in this video can benefit from really wide field of view (18-24mm). The other thing why I decided to go for a zoom was that when you switch your lenses to often, sooner or later dust get into the camera and lens.. but honesty.. prime/fixed lenses can really make a difference in quality and make life easier. If I would scan just flat surfaces form the same distance I would probably go with prime/fixed lens. But I would say.. the best lens is the one you have. What you need to do is to use the advantages it has and mitigate cons. Just use your camera settings wisely and knowledgable and you should be fine. I hope I didnt overcomplicated the answer and that answers the question :D. Cheers!
@@GrzegorzBaranArt Read the entire comment. I have a Canon 18-55 lens that is very weak and I'm not happy with it. There is also Helios 44 but it has a very small focal length. Helios 44 gives a good picture, but it is very difficult to aim, and 18-55 constantly gives an overexposed picture and I do not like the image quality. I want to buy a simple autofocus lens now, I think 18-135 will do. Thank you for such a detailed answer.
@@TheBlackFoxMaster Getting under or over exposed images has nothing with lens but the camera setting. I would say that 18-55mm lens you have is a very good lens for surface scanning, you just need to spend a bit more time with it to learn hot to get the best from it. As I mentioned its all about camera setting. Of course it depends what are you planning to scan but 18-135 wont make your life easier at all. I guess it might beneficial if I record a video about camera setting one day and explain each in details. I suggest to shot in AV mode and set the initial apreture around 8. You can set it higher it even to 11 or 16 if you use a tripod for the camera stabilisation. 8 is fine if you use at least a monopod. If you shot handheld.. I would still go with 8 but would increase the ISO. Hope that helps. Cheers!
Привет. Спасибо большое за урок в двойне спасибо за сабстанс - прям ваще респект!!!! Но я по-другому поводу! Можешь объяснить, т.е. записать урок, как выровнять фотограмму в RC-ке к примеру по полу, грубо говоря: есть пол/асфальт/дорога. Взять несколько точек с поверхности пола и указать RC-ке что это пол, т.е. чтобы она выровняла его относительно поверхности, ну и одну точку указать быть в нулях. С одной точкой в нулях понял и делал, но никак не пойму как другим точкам задать чтобы фотограмма выровнялась. Почему то ломается всё! :/ Я немного занимаюсь трекингом и делаю сканы локаций. А фотограмму я равняю через Define Ground Plate на глаз глядя через виды Top, Left и тд. Как то хотелось бы автоматизировать! :)
How do you deal with inconsistent lighting conditions? Even when shooting in the shade you can spot, after the photogrammetry is finished, a clear light gradient coming from one direction.
Hey Boris, there are many ways to deal with this. There are dedicated, often AI based delighting tools, there is a delighting tool Metashape offers. You can also consider image tweaking with the photo editing software before any images are even processed by photogrammetry softwar. You can even create your own tools like the one I designed in Substance Designer and covered as an optional deligting workflow in one of my videos already: th-cam.com/video/6pyK57fbF8c/w-d-xo.html Or you can use dedicated equipment like flash lights or light screens and illuminante the surface by yourself on your own conditions... which by the way is going to be the subject of my new incomming video :) .. about scanning with the flash, so stay tuned. But in my opinion the best way is to combine all these techniques together and get the best from each depending on the subject of capture. Of course each approach has some pros and cons, limitations and dependencies :D Hope that helps, Cheers!
Cześć. Pozwolę sobie na komentarz po polsku :). Od dość dawna oglądam Twoje materiały z ciekawością. Nie będę ukrywać, że jestem "bardzo-nie-pro", ale raczkuję w fotogrametrii... dużo zagadnień nie rozumiem jeszcze, ale w tym filmie podajesz dane związane z czasem w jakich 3DF przelicza różne etapy pracy. No i tu chciałbym się dowiedzieć jakim komputerem dysponujesz (dysponowałeś, bo to materiał sprzed roku). 3DF niby ma wsparcie dla GPU, ale jakoś nie widzę by korzystał z mojego RTXa... przynajmniej nie zauważyłem na etapie tworzenia punktów. Procesor mam mega słaby/stary bo 4 rdzeniowy (z 8 wątkami) więc gdy załączam wysoką jakość (nie najwyższą!) to "mijają lata" nim przetworzy obrazy a co dopiero wygeneruje mesh. Z góry dziękuję za odpowiedź.
Czesc, nie ma sprawy. Uzywam starszego komputera - 7 letniego - i7 z karta graficzna GF1080 Ti i 64GB calkiem powolnego RAM. Nie uzywam SSD dla zapisu scen do rekonstrukcji - uzycie SSD moze przyspieszyc proces. Pozdr.
@@GrzegorzBaranArt Dziękuję za odpowiedź. Ale, którą i7 - rozumiem, że nowej generacji czyli co najmniej 16 wątków? RAMu też mam mało bo zaledwie 16GB (i to DDR3)... No nic, czeka mnie "apdejt" sprzętu. Dzięki.
@@braccianobracciano5567 i7-7700 4.2GHz 4cores with 8 logical processors. GF1080 Ti ktorego uzywam ma tez na pokladzie 11GB RAM wiec teoretycznie moj komputer ma 64GB+11GB. Mysle ze to co w twoim przypadku moze zwalniac rekonstrukcje to wlasnie zbyt mala ilosc pamieci RAM. Jesli brakuje RAM, program zrzuca czesc informacji na dysk twardy i przerzuca ta infromacje w obydwie strony. Nawet najwolniejszy RAM jest duzo szybszy od najszybszego dysku twardego. 32GB RAM uwazane jest za wystarczajace do wieszkosci zadan. 64GB adekwatnie pozwala na przetwarzanie jeszcze wiekszej ilosci danych. Natomiast wiele zalezy od tego jak skoplikowany jest przedniot twojej rekonstrukcji i jak duzo informacji (ilosc zdjec i ich rozdzielczosc) jest przetwarzane. Przy zbyt malej ilosci RAM pomoc w tym moze trzymanie projektu na szybkim dysku SSD.. natomiast nie w takim stopniu jak dodanie RAM.
Hey Andy, and to answer your question.. no. To simplify the sharpness of the photo is mostly dependent on a camera stabilisation and its depth of field (aperture). I quess it might be worth to create a separate video about this as this is a very interesting and quite lage topic. Full frame lets more light to get to the camera sensor what in result allows you to decrease an exposure time. So you should get better images when you shot with a full frame when compared to non full frame sensor sizes. So technically if you use the same lens, and shot in same lighting conditions, and stabilise the camera on a tripod .. and set the camera the same way, you will get the same sharpness level but the image taken with a full frame sensor will need less exposure time. The problem with sharpness and image quality appears when you dont shot in labolatory and there are additional factors like low lighting, wind causing camera microshakes, or lack of stabilisation when you stabilise the camera with your hands, or a monopod. In these conditions its better to have full frame as there is a bigger chance that images taken with a full frame camera will be sharper. But if you shot in good lighting conditions, have decent stabilisation etc. it doesnt matter. Just bare in mind that even small camera movement during the exposure time might blur the image. So when you shot 6000x4000px and lose blur just two corresponding pixels you get the image of the same quality as you would shout with 3000x2000px. If the shake is larger you can lose 3 or 4.. and these are things which can really affect the reconstruction quality. Also the larger resolution might become a handicap for data processing, so better image quality let photogrammetry software to efficiently use all processing power on actual data to process. So at the end, a sharp low resolution image might have more data to the large resolution blurred one and in result give you way better 3D reconstruction output. So.. I would say, the skill mattars the most, equipment gives you some compensation. Full frame camera is better to cropped one, a DLSR is better to mobile one etc. but I saw crappy results taken with 10k£ worth cameras and masterpieces taken with mobile :). Hope that helps, dunno if I overcomplicated the answer :/ sorry if I did.. this is why I think it can be beneficial to everyone to create a decent video which covers this subject :)
@@andyjoy7391 It takes me about a month to make a single video, and have a few in 2DO queue already :). I wouldnt expect this video quick, this is why I gave you such detailed answer. A full frame camera is always better to a cropped one but the difference isnt very big and can be compensated with proper setting and supportive equipment.. like a tripod etc. So skill and good understanding of how your camera works is more relevant to the final result that size of the sensor. But if you have the same additional equipment and same skill and knowledge.. the full frame camera will give you slightly better result.
Was planning to make a video and use this case for comparison. Still might make one if you think it might be useful. Unfortunately there is no simple answer to that. Single image based reconstruction still can't compete with the photogrammetry quality, especially with such complex surfaces like this one, also I don't know any tool that can tile texture like this one better to manual seam removal. Alchemist inst any exception here .. but it isn't as bad too as Alchemist cad do a few things better. Current version of Alchemist has a really good delighting tool for albedo, probably the best in the market. Alchemist can be also used to generate roughness etc. So if you aim in quality I would say that you need to get what's the best from photogrammetry workflow and eventually combine it wit Alchemists AI. But there is significant quality gap between the result you get with the presented here workflow and what you can get with the Alchemist. Yeah.. might be worth to make a video to cover it :). Cheers!
@@GrzegorzBaranArt Really appreciates your long reply. I think the same, that the alchemist should do the job but the quality will be behind photogrammetry texture. So it depends on what we want to do. Regarding removing seam, I see that Quixel Mixer has a feature for that - no need to use Painter and Designer. But I doubt this tool can replace the other two software?
I covered it already in my photogrammetry book (gum.co/YanD) - planning to cover it even better in a new one which I plan to relase hopefully pretty soon.. but I also plan to cover it in my incoming video as its quite complex but also interesting subject. Basically its useful for color calibration as a color reference.
Mala rzecz, jak okreslasz role tekstur w Painterze zaraz po imporcie, zaznacz wszystkie i zmien role dla jednej - wybor zostanie nalozony na pozostale.
Sooooo much work involved. Thanks for sharing.
Anytime :) btw. 3DF Zephyr got much better since the time I released this video and is able to generate much denser meshes with better level of micro details. But the workflow is the same.. just saying as re-tested it recently to see the progress it made.
Amazing 🔥
From INDONESIA 🇲🇨🇲🇨
Cheers!
Bravo, excellent work! Viva clone tool!
Cheers! :D
Really appreciate the free workflow videos, as someone who can't afford to buy your photogrammetry pdf yet these a great teasers and ways to learn.
I am really happy to know you found them useful, and big thanks for watching :). Cheers!
Awesome work yet again! Wow, that location is a gold mine for texture resources too!
Thanks Nathan. Yeah, its a really good place for industrial related content :). Cheers!
awesome! thank you for sharing!
My pleasure :)
Very very in depth tutorial with explanation of why and what for each step.
Too bad that they are not all free tools but I did get a lot out it.
Thank you for the effort.
Thanks and you are welcome :). Regarding the tools used.. very often you get what you pay for. In my opinion free solutions are great to start with, but at some point most are not good enough, lacks updates, new features, stability, performance, support, quality etc. Imo the best solution is to purchase perpetual license of the software you are planning to use in long term.. even if it costs a bit more then a subscription .. this way you wont end up without any software when you stop paying. Also what I present is an optional workflow.. I would say that there are key tools you have to pay for sooner or later.. and tools which can be replaced which its free alternative. Its good to know them all as this way you can consciously decide what workflow suits you the best :) Cheers!
@@GrzegorzBaranArt I totally agree with you.
Specially in buying perpetual in stead of subscription.
I won't be getting any of this at this moment but the work flow can be used differently with other software but with more steps and less quality for a beginner
I dont have enough hours in the day to learn all this! Wish i did.
This is why I make my videos so you can save some time I had to spend.. and this is why it takes me so long to make each as sometimes.. a single simple sentence cost me weeks of actual research :D
@@GrzegorzBaranArt you are a good man for doing it! really appreciate it!
Thanks to share your experience! That was useful especially with the Zephyr settings!
My pleasure Geoffrey :). I am really happy to know that you found this video useful :)
Thanks for video!
You are welcome Vlad, and big thanks for watching :)
!!!!! great.
Cheers Jonathan :)
Thank you! this is extremely useful!
I am really glad to hear that :). Thanks for watching Juan. Cheers!
Great work! Thanks mate
Cheers! You are welcome :D
really great tutorial ! Thank you so much!
Thanks man :)
Man, that's nice one! I am from Ukraine though, i have evacuated to west part of it, and there so many surfaces and nature objects to scan!! Thank you for that workflow ❤️😀
You are welcome. Be safe
Very good!!!
Cheers man :)
Very interesting and detailed video. I especially liked seeing cars driving by, they must be wondering why some guy is photographing a brick wall :)
Thanks Barny, I am sure they did as some used horn to let me know about this :)
@@GrzegorzBaranArt congrats for 3k subs 😁😁😁
@@Barnyz Thanks.. but as I see its still just 2999 ;P
@@GrzegorzBaranArt says "3k" for me. I'm on a tablet so maybe it rounds it up 😁😁😁
Great content. Just got into photogrametry myselft. This video helps a lot!
Thanks, glad to hear that Alexander :)
Great tutorial, very helpful to me, because I am learning photogrammetry, thanks!
Glad to hear that Francis and good luck with scanning :)
Great tutorial like always!
Thank you Dimitrios, very appreciated!!! :)
Thank you
You are welcome :)
Great timing. A note, B&H has the Godox Ring Flash for $50 off, which I just bought and will test out.
So I thought it would be fun, right, to try and find the exact wall on Google Street View. Y'know, not really any particular reason. A bit of the geoguessrerer in me.
I've spent TWO HOURS walking around South Shields. I'm not even from South Shields. I've never been! I tried to find the landmarks, find the building supplies company that the sign is pointing to (next right), I had narrowed it down to the exact area. But the wall wasn't there! I kept closing Google Maps thinking "you're going mad, please stop", but the urge to find the f'king wall was taking over. Finally I had a lightbulb moment... maybe the wall doesn't exist anymore?
I put street view into "back in time" mode, went back a few years, and suddenly, in that curve of road that now stands nothing but empty sky, the wall! The building with the green trim. Those billboards. It was a suddenly breathtaking moment of clarity. But then of sadness, because that beautiful brickwork is sadly no more... But then another moment of joy, when I realised it's been forever immortalised as a PBR texture. So thank you, for letting that wall live on forever more.
goo.gl/maps/5oTwhbnPF5TSDKDB7
Its an amazing story :D and I am impressed you managed to find the exact spot :). Ive just checked on google maps and you are right, apparently part of this exact wall with the building was removed recently. To be honest I have realised that its been over a year since the last time I was there in person. Cheers!
Thanks!
You are welcome
Great man !
Cheers Lorenzo :)
Hey, awesome work and awesome tut. You mentioned you have a plane with a special UV, can you explain a little more about it? how you did and what for? Is it just scaling up the plane UVs until it's greater than the 0-1 space?
Hey, thanks a lot. Nah, the plane I made covers additional 10% for each side outside of the standard 1:1 coverage. It allows to remove side connection points coming from the offset which would be super hard when simple 1:1 plane with 1:1 uvmapping is being used. I described all the details of this plane in my ebook ( www.artstation.com/artwork/K0PYG ) and added this plane to it for those who might use it. I think I also covered quick preview for it in one of my videos but cant remember which one.
Love the tutorials Grzegorz. including seagulls :)
I'm currently working on a sandstone Priory (church) and would love to to see how you'd approach old building materials and surfaces. I've not been given drone approval yet on the building but may be able to next spring. (Malvern Priory)
Thanks Darran.. especially for watching till the end. Regarding building scanning, it depends on the size of the height and size of the object but it is totally doable. With the drone it would be just very easy and can save a lot of time and pain :) compared. Without the drone .. unless it is not Notre Dame .. I would consider to use similar technique to presented in this video.. just instead of 2m long monopod and cable remote I would use wireless remote and 'Telescopic Monopod Camera Pole'. You can get even a 10m and loner one.
Yeah, would b e cool to make a video to cover the process :). Cheers!
@@GrzegorzBaranArt Thank you so much for your response. Please, please do consider making a tutorial, it sincerely does help us non pro's to start the daunting process.. also The Witcher, i bet you have a story or two of that process :)
Ahh yessss, seagulls and drone footage :>
What about surfaces that have really big elements? Like brick wall created from 50cm big bricks or that waved metal sheet from video end? Most of time workflow will be the same, but make it tillable would be harder?
Yeah. Its hard to tile larger elements and depending on the usage, at some point larger element should be driven by the geometry itself. So for example when you scan a cliff wall and planning to use it in a game etc. you should build a complex geometry but cover it with a smaller tiling material. To keep surface interesting you can use mask and mix a few materials together. Dunno if that answer your question :).
Hi Gregorz nice video. Great like always. Thanks for share your knowlege i appreciate it so well. At this time i m still working in the Bark Texture Baking for my own personal tree creation. When i finish my work on that i hope share it with you to know your opinion about it, for me is very important. Cheers..!!
Thanks, no probs Carlos, just let me know when its ready.
you should do aerotriangulation alignment before processing photos, then replace the original photos with those processed photos for reconstruction texturing
Hey, thanks for the tip, but could you please tell me why do you believe I should? This isnt a large aerial capture with inconsistent lighting conditions etc. I usually dont have any problems with image alignment as for materials I get very even image and coverage control points distribution, consistent lighting, consistent angle and distance to the surface. So I cant really see any reason to complicate the existing process.
amazing! thanks man
Cheers Gabriel!
What if u want to use the godox ar400
What would u use instead?
A tripod?
Or something else
Probably tripod as AR400 is quite heavy. But it all depends on case. If I scan a tree trunk I might go with monopod or just hold it with bare hands
I was always interrested in how to make accurate roughness map. It looks like behind that is something more than photogrametry workflow. For me it is one big mystery. Of course there is a way by mixing all other materials into one, but it's not accurate and i'm trying to find the right way.
Yeah, roungness is very tricky since it is a kind of PBR hack. I plan to cover it in series of videos. I started them but didnt have time to finish any yet.
Great videos, very informative. Thanks for your time and effort. Btw can you please make a video comparison between Agisoft, 3DF Zephyr and RealityCapture with the same photos ?
Hey, thanks. Yes, I have it planned and even created one, but never released in form of any video :) because I decided that it is worth to wait for 1.7 update of the Metashape which is going to improve medium and high frequency noise reconstruction, and also for an update from 3DF Zephyr which is going to bring huge improvement for med and high frequency reconstruction for photoconsistency pass. I was also planning to include MeshRoom in this comparison :). So since incoming updates are going to change a lot I decided to wait.
Currently, Metashape and Reality Capture delivers similar resut.. its hard to judge which one is actually better. 3DF Zephyr fails on med and high frequency details. Metashape is the slowest one tho, 3DZephyr and RC do the job quicker.
@@GrzegorzBaranArt Thanks a lot and good to hear. I have zero knowledge of Metashape unfortunately but was counting on RealityCapture. Hoping to see upcoming video when the updates come.
Honestly fella Ive been watching a few of your videos now and the amount of effort you put in is incredible!!! I really thank you for the insight and inspiration you provide to me for the field of photogrammetry... Thank you very much. I would like to ask a question. How do you manage your files? I think it would make a great video in itself. How to be organised! Thank you again.
Thank you very much :) doing my best.
Regarding the file structure, its definitelly a very good idea for a quick video :) but I guess it depends on preferences so thought it might be too obvious.
Basically I have 2 main folders. One is called - SCANS and another one TRIPS. When I finish any capture trip I create a NUMBER-TRIPNAME folder (number is a consecutive number) in TRIP and throw everything from this trip there. In this folder I create subfolders with the gear I am copying from name - for example Canon80D or Drone or DJIOsmoPocket or Mobile etc. DjiOsmoPocket and Mobile usually contain documentary files. I take a lot of these as references but also as B-ROlls for my videos or shots I can use in my incoming book.
Next when I want to process exact scan I create a folder with its name and consequtive number in SCANS folder into a RAW subfolder.
When I process them in photoediting software I drop them into another folder called .. lets say PROCESSED_TIFFS.
Next I process scan and store the main file there. When I export scans and textures I store them into EXPORT subfolder. I bake into a BAKE subforlder. I also have subforlder with documentaries I share on my Artstation profile. When the material is ready, I save it into a subfolder with the texture name. So to give you an example:
D:/TRIPS/
--/ 101-CoastStonesAndGravel/
-----/Canon80D/
-----/OsmoPocket/
-----/Drone/
-----/Mobile/
--/ 102-DockyardConcreteAndGravel/
-----/Canon80D/
-----/OsmoPocket/
-----/Mobile/
--/ 103-ForestTrip-GrassAndTrees/
-----/Canon80D/
-----/OsmoPocket/
-----/Mobile/
D:/SCANS/
--/ 432-Gravel/
-----/raw/
-----/docs/
-----/tiff16bitsCalibrated/
-----/bakes/
-----/gb_gravel432/
-----/export/
-----/gravel432_scan.xx (photogrammetry_project)
--/ 433-Gravel/
-----/raw/
-----/docs/
-----/tiff16bitsCalibrated/
-----/bakes/
-----/gb_gravel433/
-----/export/
-----/gravel432_scan.xx (photogrammetry_project)
--/ 434-Rock/
-----/raw/
-----/docs/
-----/tiff16bitsCalibrated/
-----/bakes/
-----/gb_rock434/
-----/export/
-----/rock434_scan.xx (photogrammetry_project)
etc.
I keep this structure well organise since I backup all these files from time to time as photogrammetry related files takes a lot of space.
So I have HDDs which stores scans from 1-133, next 134-230 etc.. and similar to trips 1-30, another HDD has 31-42 another next and so on.
Hope that makes sense :)
Looks like I dont need to make a video to cover this anymore.
Cheers!
@@GrzegorzBaranArt Superstar fella!! Most Kind. Thank you for such a well considered response, that I really didn't expect to receive! So very Grateful... My organisation, in honesty, is my (but not my only) weakness... and I will be correcting this. I have been interested in 3d Design and photogrammetry at a very novice level for about 10 years... And at last I may have a nice project to begin my own portfolio, we will see. Thank you again.
Merci !
Mon plaisir :)
Amazing... ;) 4 SURE !
Cheers!
You are amazing as always mate!
Thank you very much , appreciated
Отличное качество съемки и доступное объяснение! Желаю удачи Вашему каналу.
Большое спасибо. Желаю и вам всего наилучшего. Оставайся в безопасности
perfect tut! thanks!
I just wish they implemented some easier and more precise simplify/retopo solution directly to Zephyr/RC to avoid the high number of post-softwares we need...
I totally agree. So far you can use simple decimation but it often sucks with more complex meshes. Also to make a texture you need the canvas to be as square as possible to avoid space waste. Also I guess as it looks straight forward for flat surfaces it might be an isue for complex 3D meshes. But I totally agree that simple lowpoly retopology tool with the option to align/snap to the heavy mesh might be very helpful.
I'ld use a lens like a 50mm or longer to avoid distotions which a focal lentgth like 24mm has.
I wouldnt use 50mm at all for capture like this one.. maybe35mm would also do the job.. Especially not from such close distance. I would say.. if I shot from a larger distance, and the surface is less complex and I use the tripod for camera stabilisation 50mm or even longer makes sense. But since I shot the surface from really close distance, the surface had a lot of depth and I used a monopod handled in a way you can see on this video I prioritised a shutter speed and camera's beam view over frontal data density. Basically with the 24mm mm more light gets into the camera and the shutter speed needs less time to take an image so the final images is sharper. With the 50mm I would get more blured images and less side coverage. The data captured for this reconstruction and the image coverage is way beyond of what I need for reconstruction anyway, so setting used was totally correct. Bare in mind that camera distortion doesnt matter if you know how to use it. Its just a tool and has its own pros and cons. I would agree that lens set to its edge values brings some additional consistency distortion.. but it is still ok for this type of capture. I aimed to create 4k texture set from it.. THe data I captured is still enough to create 8 or maybe even 16k. Cheers!
Very interesting as always! Is the primary reason for the switch from Metashape to 3DF Zephyr because you can get more fine detail on the mesh? How do you think the two packages compare?
I didnt switch to 3DF Zephyr :), I am just trying to play with different tools to present different workflows.. ther pros and cons. Regarding the comparison.. I am planning to make a video about this and compare Meshlab, Reality Capture, Metashape and 3DF Zephyr. Planning to do this after Metashape will sort issues with depthmaps and release 1.7 but also after 3DFlow guys finish the improvement for the Photoconsistency pass .. hopefully with the next update. So currently when I compare the software I think that Reality Capture gives the best results but is the most expensive and the most impredictable. If I score software from 1 to 10 I would say that RC is 9. Metashape is 8 but incoming updates are going to push it to 9 too. 3DF Zephyr is currently 6. After a new update they should jump to 7... maybe 7.5. And they are also very affordable.
Bare in mind that 3DF Zephyr evolves rapidly.. every change is a big change. RC introduces just super small.. slight changes .. so they basically dont move up anymore. Metashape grows faster to RC and they might beat them soon. Regarding the price and sale model.. I would say that Metashape with the incoming updates and CUDA support is going to be the best and the most affordable piece of photogrammetry software on the market :). 3DZephyr is moving really fast but its still signifficantly behind.
So if I have to pick between the Metashape and 3DZephyr.. Metashape definitely wins.. but its good to know many tools as it gives you option of consious choice.. and bare in mind that all these tools are in constant change. Something which is decent today.. might become the best option tomorrow :). So you should always pick what suits you the best.. and keep an eye to other options as everything changes very fast. 3DZephyr is a very solid piece of software. Their software is build on a very strong and solid coding base which is hopefully going to pay back with the stability while growin up .. while other software might get stuck at the same time as every new feature might hit the stability as is hard to imlement into a messy code.. just guessing to give the point.. as I said in my video.. there is no 'the only one' or 'the best tool'. All are in constant change.. and you should pick the one which suits you the best. Hope that makes sense, Cheers! :D
Dmt.. didnt know if I answered your question.. today I would pick the Metashape.. but would keep an eye on 3DZephyr.. now I did answer :)
@@GrzegorzBaranArt That's got to be the longest and most thorough reply I've ever seen on TH-cam! 😁 You're totally right about always keeping an eye on a range of software so we can always keep improving. Inspirational!
What lens did you use?
It depends. I have two lenses. For this certain and most of my recent captures I use Canon EF-S 18-135mm Nano USM lens. I used to use an Aspherical Sigma 24-70mm EX which is bigger when compared and gets brighter and sharper images with less distortion. But since I have found that Sigma is a bit to heavy, it focuses a bit slower when compared with Canon and when facing down due to its own weight the gravity untwists the focal length ring way to easily during the capture I decided that darker but lighter and better made canon zoom works better for me
Bare in mind that Sigma I mentioned is a full frame lens while the camera I use has cropped sensor. So this 24-70mm is cropped into smaller sensor.
Anyway, I was considering to get 24-70mm Canon instead but it seems to be a bit too expensive for me. The Sigma lens due to its size is tricky to be used together with the ring flash light, especially that zooming extends the length of the lens. When used with the AR400 flash light, if not perfectly aligned, extending focal tube can push the polarisation filter from the flash light off :)
I didnt cover scanning with the flash light in my videos yet but when I do I need to remember to show the issues large lens can cause.
I was thinking to use a prime lens since they are better in every way? But because zoom lens gives me more options during scanning, for example complex surfaces usually need more coverage and higher focal lenth (35-55mm) while simple and flat surfaces can be captured quicker with wider fov (24-35mm). Some surfaces with hard acces might benefit from a telephoto zoom (70-200mm) while shoting on very short distance like the brick wall in this video can benefit from really wide field of view (18-24mm).
The other thing why I decided to go for a zoom was that when you switch your lenses to often, sooner or later dust get into the camera and lens.. but honesty.. prime/fixed lenses can really make a difference in quality and make life easier. If I would scan just flat surfaces form the same distance I would probably go with prime/fixed lens.
But I would say.. the best lens is the one you have. What you need to do is to use the advantages it has and mitigate cons. Just use your camera settings wisely and knowledgable and you should be fine. I hope I didnt overcomplicated the answer and that answers the question :D. Cheers!
@@GrzegorzBaranArt Read the entire comment. I have a Canon 18-55 lens that is very weak and I'm not happy with it. There is also Helios 44 but it has a very small focal length. Helios 44 gives a good picture, but it is very difficult to aim, and 18-55 constantly gives an overexposed picture and I do not like the image quality. I want to buy a simple autofocus lens now, I think 18-135 will do.
Thank you for such a detailed answer.
@@TheBlackFoxMaster Getting under or over exposed images has nothing with lens but the camera setting. I would say that 18-55mm lens you have is a very good lens for surface scanning, you just need to spend a bit more time with it to learn hot to get the best from it. As I mentioned its all about camera setting. Of course it depends what are you planning to scan but 18-135 wont make your life easier at all. I guess it might beneficial if I record a video about camera setting one day and explain each in details. I suggest to shot in AV mode and set the initial apreture around 8. You can set it higher it even to 11 or 16 if you use a tripod for the camera stabilisation. 8 is fine if you use at least a monopod. If you shot handheld.. I would still go with 8 but would increase the ISO. Hope that helps. Cheers!
@@GrzegorzBaranArt Will spend a bit more time with it to lern hot to get the best from it in future. THX!
Привет. Спасибо большое за урок в двойне спасибо за сабстанс - прям ваще респект!!!! Но я по-другому поводу! Можешь объяснить, т.е. записать урок, как выровнять фотограмму в RC-ке к примеру по полу, грубо говоря: есть пол/асфальт/дорога. Взять несколько точек с поверхности пола и указать RC-ке что это пол, т.е. чтобы она выровняла его относительно поверхности, ну и одну точку указать быть в нулях. С одной точкой в нулях понял и делал, но никак не пойму как другим точкам задать чтобы фотограмма выровнялась. Почему то ломается всё! :/ Я немного занимаюсь трекингом и делаю сканы локаций. А фотограмму я равняю через Define Ground Plate на глаз глядя через виды Top, Left и тд. Как то хотелось бы автоматизировать! :)
How do you deal with inconsistent lighting conditions? Even when shooting in the shade you can spot, after the photogrammetry is finished, a clear light gradient coming from one direction.
Hey Boris, there are many ways to deal with this. There are dedicated, often AI based delighting tools, there is a delighting tool Metashape offers. You can also consider image tweaking with the photo editing software before any images are even processed by photogrammetry softwar. You can even create your own tools like the one I designed in Substance Designer and covered as an optional deligting workflow in one of my videos already: th-cam.com/video/6pyK57fbF8c/w-d-xo.html
Or you can use dedicated equipment like flash lights or light screens and illuminante the surface by yourself on your own conditions... which by the way is going to be the subject of my new incomming video :) .. about scanning with the flash, so stay tuned. But in my opinion the best way is to combine all these techniques together and get the best from each depending on the subject of capture. Of course each approach has some pros and cons, limitations and dependencies :D Hope that helps, Cheers!
Cześć.
Pozwolę sobie na komentarz po polsku :).
Od dość dawna oglądam Twoje materiały z ciekawością. Nie będę ukrywać, że jestem "bardzo-nie-pro", ale raczkuję w fotogrametrii... dużo zagadnień nie rozumiem jeszcze, ale w tym filmie podajesz dane związane z czasem w jakich 3DF przelicza różne etapy pracy. No i tu chciałbym się dowiedzieć jakim komputerem dysponujesz (dysponowałeś, bo to materiał sprzed roku). 3DF niby ma wsparcie dla GPU, ale jakoś nie widzę by korzystał z mojego RTXa... przynajmniej nie zauważyłem na etapie tworzenia punktów. Procesor mam mega słaby/stary bo 4 rdzeniowy (z 8 wątkami) więc gdy załączam wysoką jakość (nie najwyższą!) to "mijają lata" nim przetworzy obrazy a co dopiero wygeneruje mesh.
Z góry dziękuję za odpowiedź.
Czesc, nie ma sprawy. Uzywam starszego komputera - 7 letniego - i7 z karta graficzna GF1080 Ti i 64GB calkiem powolnego RAM. Nie uzywam SSD dla zapisu scen do rekonstrukcji - uzycie SSD moze przyspieszyc proces. Pozdr.
@@GrzegorzBaranArt Dziękuję za odpowiedź.
Ale, którą i7 - rozumiem, że nowej generacji czyli co najmniej 16 wątków? RAMu też mam mało bo zaledwie 16GB (i to DDR3)...
No nic, czeka mnie "apdejt" sprzętu. Dzięki.
@@braccianobracciano5567 i7-7700 4.2GHz 4cores with 8 logical processors. GF1080 Ti ktorego uzywam ma tez na pokladzie 11GB RAM wiec teoretycznie moj komputer ma 64GB+11GB. Mysle ze to co w twoim przypadku moze zwalniac rekonstrukcje to wlasnie zbyt mala ilosc pamieci RAM. Jesli brakuje RAM, program zrzuca czesc informacji na dysk twardy i przerzuca ta infromacje w obydwie strony. Nawet najwolniejszy RAM jest duzo szybszy od najszybszego dysku twardego. 32GB RAM uwazane jest za wystarczajace do wieszkosci zadan. 64GB adekwatnie pozwala na przetwarzanie jeszcze wiekszej ilosci danych. Natomiast wiele zalezy od tego jak skoplikowany jest przedniot twojej rekonstrukcji i jak duzo informacji (ilosc zdjec i ich rozdzielczosc) jest przetwarzane. Przy zbyt malej ilosci RAM pomoc w tym moze trzymanie projektu na szybkim dysku SSD.. natomiast nie w takim stopniu jak dodanie RAM.
@@GrzegorzBaranArt Dziękuję. Będę mieć na uwadze wszystkie te informacje.
Hello, isn’t the sharpness of a photo dictated by shooting with a full camera or a half camera?
Hey Andy, and to answer your question.. no. To simplify the sharpness of the photo is mostly dependent on a camera stabilisation and its depth of field (aperture). I quess it might be worth to create a separate video about this as this is a very interesting and quite lage topic. Full frame lets more light to get to the camera sensor what in result allows you to decrease an exposure time. So you should get better images when you shot with a full frame when compared to non full frame sensor sizes.
So technically if you use the same lens, and shot in same lighting conditions, and stabilise the camera on a tripod .. and set the camera the same way, you will get the same sharpness level but the image taken with a full frame sensor will need less exposure time. The problem with sharpness and image quality appears when you dont shot in labolatory and there are additional factors like low lighting, wind causing camera microshakes, or lack of stabilisation when you stabilise the camera with your hands, or a monopod. In these conditions its better to have full frame as there is a bigger chance that images taken with a full frame camera will be sharper. But if you shot in good lighting conditions, have decent stabilisation etc. it doesnt matter. Just bare in mind that even small camera movement during the exposure time might blur the image. So when you shot 6000x4000px and lose blur just two corresponding pixels you get the image of the same quality as you would shout with 3000x2000px. If the shake is larger you can lose 3 or 4.. and these are things which can really affect the reconstruction quality. Also the larger resolution might become a handicap for data processing, so better image quality let photogrammetry software to efficiently use all processing power on actual data to process.
So at the end, a sharp low resolution image might have more data to the large resolution blurred one and in result give you way better 3D reconstruction output.
So.. I would say, the skill mattars the most, equipment gives you some compensation. Full frame camera is better to cropped one, a DLSR is better to mobile one etc. but I saw crappy results taken with 10k£ worth cameras and masterpieces taken with mobile :).
Hope that helps, dunno if I overcomplicated the answer :/ sorry if I did.. this is why I think it can be beneficial to everyone to create a decent video which covers this subject :)
@@GrzegorzBaranArt Thank you very much for your reply and look forward to your video
@@GrzegorzBaranArt Hello I am looking forward to the teacher’s video, I don’t know when the video will be released
@@andyjoy7391 It takes me about a month to make a single video, and have a few in 2DO queue already :). I wouldnt expect this video quick, this is why I gave you such detailed answer. A full frame camera is always better to a cropped one but the difference isnt very big and can be compensated with proper setting and supportive equipment.. like a tripod etc. So skill and good understanding of how your camera works is more relevant to the final result that size of the sensor. But if you have the same additional equipment and same skill and knowledge.. the full frame camera will give you slightly better result.
@@GrzegorzBaranArt Thank you very much
Hi, how is this compare to Alchemist AI?
Was planning to make a video and use this case for comparison. Still might make one if you think it might be useful. Unfortunately there is no simple answer to that. Single image based reconstruction still can't compete with the photogrammetry quality, especially with such complex surfaces like this one, also I don't know any tool that can tile texture like this one better to manual seam removal. Alchemist inst any exception here .. but it isn't as bad too as Alchemist cad do a few things better. Current version of Alchemist has a really good delighting tool for albedo, probably the best in the market. Alchemist can be also used to generate roughness etc. So if you aim in quality I would say that you need to get what's the best from photogrammetry workflow and eventually combine it wit Alchemists AI. But there is significant quality gap between the result you get with the presented here workflow and what you can get with the Alchemist. Yeah.. might be worth to make a video to cover it :). Cheers!
@@GrzegorzBaranArt Really appreciates your long reply. I think the same, that the alchemist should do the job but the quality will be behind photogrammetry texture. So it depends on what we want to do. Regarding removing seam, I see that Quixel Mixer has a feature for that - no need to use Painter and Designer. But I doubt this tool can replace the other two software?
having a hard time understanding what the color checker does , can you explain?
I covered it already in my photogrammetry book (gum.co/YanD) - planning to cover it even better in a new one which I plan to relase hopefully pretty soon.. but I also plan to cover it in my incoming video as its quite complex but also interesting subject. Basically its useful for color calibration as a color reference.
Mala rzecz, jak okreslasz role tekstur w Painterze zaraz po imporcie, zaznacz wszystkie i zmien role dla jednej - wybor zostanie nalozony na pozostale.
Wielkie dzieki za tip!!! To dosc duza rzecz :) wbrew pozorom.