Usually I do photogrammetry with stills but I tried out to capture my dog with a 4K mobile phone video, since my dog can´t hold still I needed something quick. What should I say, I was really surprised by the details and quality. It is nowhere near as detailed as stills would be but considering it took me about 1 min of video with all the compression artifacts etc. it is great. I would have never even been able to capture my dog with stills anyways. I am now thinking about borrowing a 6k camera for that reason. The great part is you always get enough paralax from a video. I would even say videos is where photogrammetry really shines. With future 6k and 8k cameras I can see great results.
i think stablization is key. I love using video frames for photogrammetry with M1 Mac xcode workflow. I airdrop iphone videos to around 300 frames and get great results
I really wanna try out photogrammetry on Apple's processors. Apparently Apple's Object capture API is extremely optimised for the M series processors. Can't wait for the new iMacs and Mac Pros to come out!
My mirrorless can capture video at 4k 60fps. I'm planning to retopologize my capture so this makes me confident video should give me some good enough results
Great videos! And veeery impressive scan results! Are you using the same setup when taking the photos now as in your first video or has it progressed/changed? I see a lot of people use ring light. Have you experimenting with that?
Glad you enjoy the videos! My setup hasn't changed that much. I have one extra light (the same brand and model as the one I already had) but everything else is the same. A ring light can work perfectly fine but I prefer bigger light sources. It helps spread out the light evenly and it also helps with hard shadows.
Excellent video, enough information to make me want to try this. Maybe it was just me , but the music, while nice, was distracting. I think maybe it should be further in the background. Not complaining just observing. Thank You...
nice video !! looks like that for 3D background environment for film (where we don't always need high quality) might be a good solution... tell me something please. Usually when we use photos we use more than one layer.. for example we take photos from 2 meters away from the object (and doing the circle) and then a new series of photos closer to 0.5m (for example) to catch textures, details etc.. in video can we do the same ? to have for example a 1st video and a 2nd one closer ?
For the exterior shots I didn't bother trying to shoot properly but usually you would have to wait for an overcast day and then tune things with an application like Agisoft's Delighter. When shooting in a controlled environment it's relatively easy to get rid of the shadows. Big diffuse lights is what you need!
I'm using photocatch (the OSX version not the iPhone app) which is a small little application taking advantage of Apple's Object capture API. But any other photogrammetry app will do. Apple's API is incredibly optimised so you can get great results really fast.
I think you exaggerating a bit the frames grab you showed is almost identical the the still, also in most cases these models are good enough unless you doing some Hollywood vfx shot
Great Video, thank you! I'm looking for user-friendly apps to capture rooms or multiple spaces in a house for showcasing in VR. (+Scenario 2: I don't have an iPhone) and I'm seeking a workflow that doesn't require programming skills. Which apps would you recommend for this purpose - for scanning and for showing in VR?
Things might be tricky without an iPhone. You basically need a phone with LiDAR and an app that can scan a space using LiDAR. There might be good options on the Android front but I can't give you any suggestions since I'm not familiar with the Android space. Alternatively you could use regular photogrammetry. Similar to what I've done in this video here: ruclips.net/video/xPkB6I30cxg/видео.html
I am impressed by your results and would like to attempt to replicate them using 360 video optimized with 120 FPS in 8K. Can you elaborate on your software, workflow and processing time? Any suggestions are appreciated. TIA
Hey there. I'm currently using Photocatch (desktop version) which is a mac app taking advantage of Apple's photogrammetry API. The processing time depends on the amount of pictures taken so it could be anywhere between 15 minutes to 40 minutes or more. If you want to see more about my process make sure to check the photogrammetry playlist on my channel. I go through a lot of the process and the different software I'm using.
Sorry if I missed it: I wonder what bit rate you used? The higher the bit rate the richer the texture in the video should be; other ideas would be raw output. I read a comment below about getting a dog with video, that is an interesting application. Great video, thanks!
Thanks! For Photogrammetry I'm using the desktop version of Photocatch and also Metashape. Both do the same thing but since Photocatch uses Apple's SDK the results are very fast.
Hi Balaji, sounds like he took a frame/photo every 24 frames of his videos. So he converted his video to a photo sequence, like .jpeg or .png sequence.... opened a photogammetry software like Meshroom which I think is free and opened the folder of his photo sequence and clicked control as he selected every 24 images from the sequence...then dragged those selected frames into the Meshroom media file box area... then you run it.
Why not simply raise the iso? a GH5 won’t be an issue at iso 3200/4000. So you can use a sharper f stop. Then after import the clip in Davinci. use auto stabilize, some noise reduction of needed. Then export the clip to a 10 bit codec like Prores, and you are good to go.
It's a fine balance game! Increasing the ISO will add noise and reduce detail while auto stabilising won't really solve the issue of camera shake and reduced detail because of it. Same applies with noise reduction. Lot's of detail will be smeared/lost. Better shoot it right the first time rather than trying to fix things in post!
@@marvelousdecay Totally agree. Its always a dance. And a learning game. Higher iso usually adds problematic noise above 4000 on bodies like a gh5 and F4 aperture for example will enable easier manual focus while moving and also compensates for sharpness you may loose at open apertures depending on the lens. So yeah its always a balance act.. And a gimbal would be also be a good idea lol.
After shooting video you directly uploaded this video in Photogrammetry Software or you have converted this video in Images.. please tell this pipeline?
Either approach is perfectly fine. It mostly depends on the capabilities of the photogrammetry software you’re using and the amount of time you want to dedicate per scan. I used the video straight out of the camera and told the software to use a frame every 15 or 29 frames or so. But you could export the video in to stills and pick the ones you want. It’s just going to be more time consuming
what software you used to "cut" the video into frames...? and how did you use individual frames (the best/sharpest) to use for photogrammetry...? you did it all "by the eye"...? ("manually")...?
There are different ways to do it. You could use after effects, final cut, premiere, quicktime etc to export the video as an image sequence and pick the frames you want. You can also let the photogrammetry software do the picking. Photocatch has an option to use video and you can set up at which intervals to grab frames. Whatever suits you best.
@@marvelousdecay thanks, but i thought that you should use only the "sharpest" frames, and theres a problem how to pick them...? sure you can do it "by the eye" but if you have hundreds or even thousands of frames, its not really effective... AFAIK Agisoft Metashape has some sort of "evaluation" feature (for the sharpness of the photos), so you could load all hundreds/thousands frames into agisoft, run the feature and than look at the number and pick the sharpest.
I wonder why there isn't software that could take photos and videos of an object, or lidar etc and use like AI or algorithms to analyze all that content and construct a 3d model thats the best approximation--perhaps the user could tweak the weight of how each source contributes etc--but i think AI 3d tools have still just barely emerged--here's hoping for better solutions !
I dont really get the motion blur argument. Shutter speed has the same effect on photos as it has on video. Also taking photos with a high shutter speed needs a lower aperture aswel, its not any different from video. You mentioned shooting stills on a tripod, don't you mean you get more blur with video because you are doing it handheld?
The video certainly doesn’t have an ad every minute. But even if it did you certainly wouldn’t be seeing all of them. If you don’t want to see ANY ads you can always get RUclips premium. Otherwise I fail to see how you can complain about a video you watch for free, having a way to compensate the creator.
Usually I do photogrammetry with stills but I tried out to capture my dog with a 4K mobile phone video, since my dog can´t hold still I needed something quick. What should I say, I was really surprised by the details and quality. It is nowhere near as detailed as stills would be but considering it took me about 1 min of video with all the compression artifacts etc. it is great. I would have never even been able to capture my dog with stills anyways. I am now thinking about borrowing a 6k camera for that reason. The great part is you always get enough paralax from a video. I would even say videos is where photogrammetry really shines. With future 6k and 8k cameras I can see great results.
Idea - shoot with the drone in cine mode which preserves detail since no color blows out the detail.
Awesome one buddy! Thanks for sharing!
i think stablization is key. I love using video frames for photogrammetry with M1 Mac xcode workflow. I airdrop iphone videos to around 300 frames and get great results
I really wanna try out photogrammetry on Apple's processors. Apparently Apple's Object capture API is extremely optimised for the M series processors. Can't wait for the new iMacs and Mac Pros to come out!
Going to try out 8k stabilized on ultra 22. Been in the field professionally for about eight years.
Thank you for your work. Waiting a new videos
great video quality, keep it up.
Oh man, you were talking down your captures but for me those are more than good enough! I gotta get out there with my camera immediately!
Do it! Photogrammetry is a lot of fun!
Thanks for the great content 👌
You’re welcome! Glad you enjoy it!
My mirrorless can capture video at 4k 60fps. I'm planning to retopologize my capture so this makes me confident video should give me some good enough results
Great videos! And veeery impressive scan results! Are you using the same setup when taking the photos now as in your first video or has it progressed/changed? I see a lot of people use ring light. Have you experimenting with that?
Glad you enjoy the videos! My setup hasn't changed that much. I have one extra light (the same brand and model as the one I already had) but everything else is the same.
A ring light can work perfectly fine but I prefer bigger light sources. It helps spread out the light evenly and it also helps with hard shadows.
Excellent video, enough information to make me want to try this. Maybe it was just me , but the music, while nice, was distracting. I think maybe it should be further in the background. Not complaining just observing. Thank You...
great video
nice video !! looks like that for 3D background environment for film (where we don't always need high quality) might be a good solution... tell me something please. Usually when we use photos we use more than one layer.. for example we take photos from 2 meters away from the object (and doing the circle) and then a new series of photos closer to 0.5m (for example) to catch textures, details etc.. in video can we do the same ? to have for example a 1st video and a 2nd one closer ?
Yep you can follow the same procedure as stills. You could just have one long video and then edit out the parts that you know won't work.
What do you do about shadows? I mean the environment shadows that are casted onto the object when photographing. Do they get baked in the texture?
For the exterior shots I didn't bother trying to shoot properly but usually you would have to wait for an overcast day and then tune things with an application like Agisoft's Delighter.
When shooting in a controlled environment it's relatively easy to get rid of the shadows. Big diffuse lights is what you need!
What software do you use for Photogrammetry? The built in stuff in Cinema 4d?
I'm using photocatch (the OSX version not the iPhone app) which is a small little application taking advantage of Apple's Object capture API. But any other photogrammetry app will do. Apple's API is incredibly optimised so you can get great results really fast.
I think you exaggerating a bit the frames grab you showed is almost identical the the still, also in most cases these models are good enough unless you doing some Hollywood vfx shot
Great Video, thank you! I'm looking for user-friendly apps to capture rooms or multiple spaces in a house for showcasing in VR. (+Scenario 2: I don't have an iPhone) and I'm seeking a workflow that doesn't require programming skills. Which apps would you recommend for this purpose - for scanning and for showing in VR?
Things might be tricky without an iPhone. You basically need a phone with LiDAR and an app that can scan a space using LiDAR. There might be good options on the Android front but I can't give you any suggestions since I'm not familiar with the Android space.
Alternatively you could use regular photogrammetry. Similar to what I've done in this video here:
ruclips.net/video/xPkB6I30cxg/видео.html
I think video would be a useful option for drone orthomosaic, especially in daylight with a fixed-wing drone.
Yep. You can do a whole lot with drone footage! I did a couple of tests, and you can get some nice 3d geometry out of it.
I am impressed by your results and would like to attempt to replicate them using 360 video optimized with 120 FPS in 8K. Can you elaborate on your software, workflow and processing time? Any suggestions are appreciated. TIA
Hey there. I'm currently using Photocatch (desktop version) which is a mac app taking advantage of Apple's photogrammetry API. The processing time depends on the amount of pictures taken so it could be anywhere between 15 minutes to 40 minutes or more.
If you want to see more about my process make sure to check the photogrammetry playlist on my channel. I go through a lot of the process and the different software I'm using.
Sorry if I missed it: I wonder what bit rate you used? The higher the bit rate the richer the texture in the video should be; other ideas would be raw output. I read a comment below about getting a dog with video, that is an interesting application. Great video, thanks!
Yeah I tried multiples settings. Even RAW footage at 400Mbps but it's still not as clear as a still image. But it's definitely good enough!
It doesn't work when the object is moving and background fixed. I tried to capture an actor fullbody and he turned around.
Love your content, what software do you use ?
Thanks! For Photogrammetry I'm using the desktop version of Photocatch and also Metashape. Both do the same thing but since Photocatch uses Apple's SDK the results are very fast.
How about using the GoPro11 for this? Wouldn't it's specs be great for this?
GoPro could work too. As long as it's not shaky footage you're good.
Nice video.... But can u please elaborate how you have done this complete procedure from shooting video to 3D model
Hi Balaji, sounds like he took a frame/photo every 24 frames of his videos. So he converted his video to a photo sequence, like .jpeg or .png sequence.... opened a photogammetry software like Meshroom which I think is free and opened the folder of his photo sequence and clicked control as he selected every 24 images from the sequence...then dragged those selected frames into the Meshroom media file box area... then you run it.
@@4rjohny5 thanks for the reply sir
what software do you use for photogrammetry
It’s called photocatch and it uses Apple’s photogrammetry API
Why not simply raise the iso? a GH5 won’t be an issue at iso 3200/4000. So you can use a sharper f stop. Then after import the clip in Davinci. use auto stabilize, some noise reduction of needed. Then export the clip to a 10 bit codec like Prores, and you are good to go.
It's a fine balance game! Increasing the ISO will add noise and reduce detail while auto stabilising won't really solve the issue of camera shake and reduced detail because of it. Same applies with noise reduction. Lot's of detail will be smeared/lost.
Better shoot it right the first time rather than trying to fix things in post!
@@marvelousdecay Totally agree. Its always a dance. And a learning game. Higher iso usually adds problematic noise above 4000 on bodies like a gh5 and F4 aperture for example will enable easier manual focus while moving and also compensates for sharpness you may loose at open apertures depending on the lens. So yeah its always a balance act.. And a gimbal would be also be a good idea lol.
hi! which programm are u using?
Photocatch for scanning, Cinema 4D for UVs and Zbrush for retopo and texture creation
After shooting video you directly uploaded this video in Photogrammetry Software or you have converted this video in Images.. please tell this pipeline?
Either approach is perfectly fine. It mostly depends on the capabilities of the photogrammetry software you’re using and the amount of time you want to dedicate per scan. I used the video straight out of the camera and told the software to use a frame every 15 or 29 frames or so. But you could export the video in to stills and pick the ones you want. It’s just going to be more time consuming
@@marvelousdecay thanks for the answer sir
Nikon Z9 or Z8 - 8K res for 4000-5500 USD
what software you used to "cut" the video into frames...? and how did you use individual frames (the best/sharpest) to use for photogrammetry...? you did it all "by the eye"...? ("manually")...?
There are different ways to do it. You could use after effects, final cut, premiere, quicktime etc to export the video as an image sequence and pick the frames you want.
You can also let the photogrammetry software do the picking. Photocatch has an option to use video and you can set up at which intervals to grab frames.
Whatever suits you best.
@@marvelousdecay thanks, but i thought that you should use only the "sharpest" frames, and theres a problem how to pick them...? sure you can do it "by the eye" but if you have hundreds or even thousands of frames, its not really effective... AFAIK Agisoft Metashape has some sort of "evaluation" feature (for the sharpness of the photos), so you could load all hundreds/thousands frames into agisoft, run the feature and than look at the number and pick the sharpest.
I wonder why there isn't software that could take photos and videos of an object, or lidar etc and use like AI or algorithms to analyze all that content and construct a 3d model thats the best approximation--perhaps the user could tweak the weight of how each source contributes etc--but i think AI 3d tools have still just barely emerged--here's hoping for better solutions !
Good vid man, I always liked morphing it’s a good kind of cheesy
I think you were meaning to post this on the morphing video but I’ll take the compliment. :D
@@marvelousdecay lol yes, but I did watch this too. It was interesting to see , I came from reddit so I thought to check out your other stuff.
Huw about blackmagic 12k :D
Zephyr ?
I dont really get the motion blur argument. Shutter speed has the same effect on photos as it has on video. Also taking photos with a high shutter speed needs a lower aperture aswel, its not any different from video. You mentioned shooting stills on a tripod, don't you mean you get more blur with video because you are doing it handheld?
because camera is moving ??
Then the " Compu Geek" fella call it " Brute Force photogrammetry " to video for photogrammetry... lol
Interesting video. (Sorry this comment is so lame.)
ads literally every minute. this vid is not about photogrammetry with ads. its an ad with some photogrammetry content in between
The video certainly doesn’t have an ad every minute. But even if it did you certainly wouldn’t be seeing all of them.
If you don’t want to see ANY ads you can always get RUclips premium. Otherwise I fail to see how you can complain about a video you watch for free, having a way to compensate the creator.