i was going to install this today but watched how automatic1111 was removed from git hub and was mad so i looked this subject of nerf and your video made my day and auto1111 ok now
Holy shit, THANKS! I was just about to start installing CMake and all that stuff again, I had totally missed this! Good thing I found this video first, thanks again!
Also... we still need to use colmap before running the instant-ngp training correct? Is there a way to do this within this new binary or are there other binaries that can do this? As it stands, I'd still need to use my conda ngp install to process the dataset with colmap first.
You still need camera pose information. There is a windows binary version of COLMAP that requires no code compiling. You can also use pose information from the Recon-3D app on an iPhone.
@@EveryPoint I have run into the COLMAP GUI by absolute chance 😅😅 I have also compiled everything fine and have made multiple NeRFs. Using conda. Just wanted to know if this would work in one go rather than 2 processes. Thanks for your videos man! Really appreciate it!
@@EveryPoint I'd love you even more if that's any encouragement 🤣🤣 It would be handy for us thinkering with NeRFs, but definitely also in general for python if it's done well. Which I have no doubt it would be.
I think what is confusing is that it is still not really a simple solution but a simpler one. However, I am trying to figure out the data process. Does COLMAP need to be installed into the data folder directory and do I still need to transfer the transforms json file into my data set before dragging into the GUI? I think it would be helpful to mention the details of the data process (thank you so much for the info btw :))
Does allowing the training time to go on for a longer period of time help reduce / get rid of floating cloud clumps? or is that a sign I needed more information from recording the area I am trying to create a nerf?
That is great! Especially for those who use DaVinci Resolve video editing software. I found that Anaconda may interfere with that editing software. So binary code means no interference. Thanks!
Interesting, we edit with DaVinci Resolve as well and have no ran into any issues. That being said, we believe you when you say there is conflicts! We use Miniconda and many manually installed packages on top of it. The newest version of Anaconda creates all sorts of issues with our legacy code.
Hey i downloaded it and it worked after installing ffmpeg path.. but now does the binary stuff can treat custom videos or pictures into nerfs ? please. Sorry i'm the noobiest noob in nerf thangs
I've installed the Luma Ai application on my Iphone 12 Pro and I've already done some amazing tests. I just downloaded and installed this application to test on my PC which does not require any code (I do not know how to code), but as I practice Photogrammetry and use a drone I am very interested in this new technology. I tested the Fox File and it works, but in what format should I save my captures to be able to load them into this application?
Check out our video where we automate that part of the pipeline. Note, you need FFMPEG and COLMAP pre installed per the GitHub repo documentation. One Click End-to-End Automated Instant NeRF ua-cam.com/video/-ummgy1Qy00/v-deo.html
Making money, starting a business and self development can only be achieved if you consistantly work on it, it is my personal experience... if you are reading this keep in mind that you cannot become successful in a week. BTW: enjoyed watching your video, thanks. a fellow creator=
has anyone been able to get this to work on a 10 series 4GB Nvidia card? my laptop has a 4GB GTX 1050 in it. the exe launches and runs fine, but the program crashes when i try to drag the fox data set in edit: it's working with other datasets, so i'm assuming it's just a vram limitation :(
@@Legnog822 it’s computationally hard to render millions of pixels from a neural network multiple times a second. GPUs are not currently architected for this application. It’s geared for the triangle mesh world.
Has this been configured to work with video images? I would like to start with video taken from different perspectives and generate a moving 3D video scene.
Yes, you are training a Neural Radiance Field (based on neural networks). You can then render 3D visualizations or compute (a rather poor quality) mesh.
Which apps create compatible files for NeRFs? I thought I would be able to simply take a bunch of photos of something and have it use regular photos for creating NeRFs - I guess that's not the case?
Cool, thank you. Is the max rendering resolution determined by the GPU? My training the fox stops at a 135*76 render resolution and does not progress any more on a NVIDIA Quattro P5000.
So cool i like it but i got an issue after install and try to drag the fox folder it brakes and sends and error out of memory cuda, any idea what am i doing wrong?
I don't understand this video. You ran the program, then dragged a folder into the black window, and BOOM, you had instant results. I get nothing, even with the fox. What are you actually doing??
Hi! Thanks for the video, but i have a problem, when I try to drop my folder to this Instant ngp. Program crashes every time with every folder, but it works great with original fox folder. Besides, i haven't any problem, if i try to use only anaconda code and of course, i made my transforms file every time. Thanks!
es increible... no hay ninguna noticia en cuanto al exportado de mallas, o nube de puntos, o algun tipo de herramienta para medir distancias dentro de nerf? en mi trabajo serviria mucho medir distancias para poder procesar escenas de choques automovilisticos. Saludos!
it is nice, but it still is too damn hard to use. I tried with pics, videos, from different angles and all and I couldn't make anything remotely as good as what you did there. this tech still need to get way better before it is usable by people like me who don't want to have to know how to perfectly take the pictures for it to work remotely good.
Congratulations on bringing this rich content to us! I'm not able to make it work in my Win 11 Ryzen 7 RTX 3050 4G The software opens and then closes. What am I doing wrong?
Woo! Looks like it's never been a better time for me to get into this stuff!
Yesss!
This video was like getting through weeks of arg into more cool activities!! Thanks tons!!
Just wished me made it a week ago when it was silently released! We skipped training the data. We are already working on a follow up video.
Glad the christmas present was well received!
Yes, thank you for the present!
❤
i was going to install this today but watched how automatic1111 was removed from git hub and was mad so i looked this subject of nerf and your video made my day and auto1111 ok now
Holy shit, THANKS! I was just about to start installing CMake and all that stuff again, I had totally missed this! Good thing I found this video first, thanks again!
This is a good option if you aren’t concerned with using Python bindings. Most creators won’t need that functionality.
i see this right after i've been trouble shooting how to set it up from an older tutorial for over an hour, i want to cry
Also... we still need to use colmap before running the instant-ngp training correct? Is there a way to do this within this new binary or are there other binaries that can do this? As it stands, I'd still need to use my conda ngp install to process the dataset with colmap first.
You still need camera pose information. There is a windows binary version of COLMAP that requires no code compiling. You can also use pose information from the Recon-3D app on an iPhone.
@@EveryPoint I have run into the COLMAP GUI by absolute chance 😅😅 I have also compiled everything fine and have made multiple NeRFs. Using conda.
Just wanted to know if this would work in one go rather than 2 processes.
Thanks for your videos man! Really appreciate it!
Perhaps I should make instructions on how to make a Python scripts that runs everything all in one go.
@@EveryPoint That would be super helpful! Would you please?
@@EveryPoint I'd love you even more if that's any encouragement 🤣🤣
It would be handy for us thinkering with NeRFs, but definitely also in general for python if it's done well. Which I have no doubt it would be.
I think what is confusing is that it is still not really a simple solution but a simpler one. However, I am trying to figure out the data process. Does COLMAP need to be installed into the data folder directory and do I still need to transfer the transforms json file into my data set before dragging into the GUI? I think it would be helpful to mention the details of the data process (thank you so much for the info btw :))
Does allowing the training time to go on for a longer period of time help reduce / get rid of floating cloud clumps? or is that a sign I needed more information from recording the area I am trying to create a nerf?
That is great! Especially for those who use DaVinci Resolve video editing software. I found that Anaconda may interfere with that editing software. So binary code means no interference. Thanks!
Interesting, we edit with DaVinci Resolve as well and have no ran into any issues. That being said, we believe you when you say there is conflicts! We use Miniconda and many manually installed packages on top of it. The newest version of Anaconda creates all sorts of issues with our legacy code.
Nice! How do you use own pictures to training with this? Can't find a way..
i have a 1660 which one should i pick
How do you make the datasets for these?
Working on an automated script for that, hold tight.
Hey i downloaded it and it worked after installing ffmpeg path.. but now does the binary stuff can treat custom videos or pictures into nerfs ? please. Sorry i'm the noobiest noob in nerf thangs
You will need to prepare custom image sets. Hold tight. We will have a no-code solution in the next few days for that as well.
@@EveryPoint it looks like i got a colmap usable in windows if it can help 😂😂🙌 thank you man
Hey. May you upload installation zip ? as it not work on GitHub now(
I've installed the Luma Ai application on my Iphone 12 Pro and I've already done some amazing tests. I just downloaded and installed this application to test on my PC which does not require any code (I do not know how to code), but as I practice Photogrammetry and use a drone I am very interested in this new technology. I tested the Fox File and it works, but in what format should I save my captures to be able to load them into this application?
looking into the same
did you figure this out?
i can't just put images in an instant-nerf window? ( i have to train it before? what is the best way to train?
Check out our video where we automate that part of the pipeline. Note, you need FFMPEG and COLMAP pre installed per the GitHub repo documentation. One Click End-to-End Automated Instant NeRF
ua-cam.com/video/-ummgy1Qy00/v-deo.html
Thanks for sharing! Very quick and informative for a newbie such as myself. :-)
Making money, starting a business and self development can only be achieved if you consistantly work on it, it is my personal experience... if you are reading this keep in mind that you cannot become successful in a week. BTW: enjoyed watching your video, thanks. a fellow creator=
has anyone been able to get this to work on a 10 series 4GB Nvidia card? my laptop has a 4GB GTX 1050 in it. the exe launches and runs fine, but the program crashes when i try to drag the fox data set in
edit: it's working with other datasets, so i'm assuming it's just a vram limitation :(
We highly recommend a 20XX or better. What is happening is that you’re running out of VRAM. It takes quite a bit.
@@EveryPoint considering i have one of the weakest cards available that's still capable of running stable diffusion i'm not even surprised
@@Legnog822 it’s computationally hard to render millions of pixels from a neural network multiple times a second. GPUs are not currently architected for this application. It’s geared for the triangle mesh world.
Has this been configured to work with video images? I would like to start with video taken from different perspectives and generate a moving 3D video scene.
Hello! Thanks for the video. How do you toggle 3d mode? I can just see the 2d photos of the fox as I drag it over...
ok, I see CUDA... and CMake? are required?
OK, downloaded CUDA and CMake. Still no 3d render...
What if you wanted export a mesh, masked out?
any way to export obj or fbx?
We suggest checking out NerfStudio for this functionality.
Great video! Any resources for trying this out on Mac?
F mac
Crashes when I drag the fox folder into the window.
What GPU are you using? Usually Instant NGP crashes when you run out of VRAM.
Same here. I have a GEFORCE 1050Ti. It just crashes as soon as I drag the fox folder to the window.
@@EveryPoint same here, im using nvidia geforce MX 110, VRAM 2010 Mb, what do you think
is there any method to extract the mesh
how to get these transform JSON files from?
When you said "training" I thought you meant going from Images to 3d. Can you do that with the binary ?
Yes, you are training a Neural Radiance Field (based on neural networks). You can then render 3D visualizations or compute (a rather poor quality) mesh.
Which apps create compatible files for NeRFs? I thought I would be able to simply take a bunch of photos of something and have it use regular photos for creating NeRFs - I guess that's not the case?
The one for Pascal doesn't even work. It complains it was compiled for type 70, which isn't even Pascal.
Ouch, good catch. I fixed the script and correct Pascal binaries (CC 61) should be up in ~1h
@@Tom94ness nice one!
Cool, thank you. Is the max rendering resolution determined by the GPU? My training the fox stops at a 135*76 render resolution and does not progress any more on a NVIDIA Quattro P5000.
how to export mesh file?
Nice, thanks for the share!
🙌
So cool i like it but i got an issue after install and try to drag the fox folder it brakes and sends and error out of memory cuda, any idea what am i doing wrong?
I don't understand this video. You ran the program, then dragged a folder into the black window, and BOOM, you had instant results. I get nothing, even with the fox. What are you actually doing??
I did that too and got direct results. you sure you are running on the right gpu?
Hi! Thanks for the video, but i have a problem, when I try to drop my folder to this Instant ngp. Program crashes every time with every folder, but it works great with original fox folder. Besides, i haven't any problem, if i try to use only anaconda code and of course, i made my transforms file every time. Thanks!
Same problem here, but also with the fox folder.
I suggest jumping on the official NVIDIA Instant-NGP Github repository discussion board for help.
@@EveryPoint Thanks, but I did not find anything about the problem there.
Awesome,
Thanks for sharing
es increible... no hay ninguna noticia en cuanto al exportado de mallas, o nube de puntos, o algun tipo de herramienta para medir distancias dentro de nerf? en mi trabajo serviria mucho medir distancias para poder procesar escenas de choques automovilisticos. Saludos!
bro how did you do the custom one with ur son?
Perfect
:)
i dont get what you dropped in. are those mp4 files? jpg files? please try and be more descriptive, as not evrryone watching is experienced. thanks
I forgot how to actually make a NERF... damn
it is nice, but it still is too damn hard to use. I tried with pics, videos, from different angles and all and I couldn't make anything remotely as good as what you did there. this tech still need to get way better before it is usable by people like me who don't want to have to know how to perfectly take the pictures for it to work remotely good.
Congratulations on bringing this rich content to us! I'm not able to make it work in my
Win 11
Ryzen 7
RTX 3050 4G
The software opens and then closes. What am I doing wrong?