Thanks you for the requirement bat! , I didn't noticed the requirement bat at first and I was shocked by so much tech talk so I have been all day installing dependencies one by one (as a 3D artist I didn't even had anaconda installed), don't be like me guys, just put the requirements bat file in the instant ngp folder and say thanks hahahaha
Thank you very much! This problem has been bothering me for a long time, but because I don't know how to program, I haven't been able to make the program run properly. I really appreciate your help!
Seeing this makes me wonder if you could run an instance of instant NGP in a cloud GPU provider like Lambda Labs. That way, you could still play use this tool even if you don't have a compatible setup. Export all necessary files to mess with in Blender or something. Not sure how much functionality might be lost though. Would be interested to hear your thoughts on this.
Great follow up video! I noticed that the windows zip Nvidia provided didn't include the required scripts for downloading COLMAP if you didn't already have it installed and tried to leave a comment about how I got around it but it seems like it got blocked. Glad you put this together to help simplify the process!
@@EveryPoint Looks like they fixed the exclusion of the "download COLMAP" scripts being missing from the binaries two days ago, so I guess that's a non-issue now. Now my struggle is to convince colmap to use CUDA rather than the CPU. Guess its time to go back an RTFM. :)
I need help how to fix this Finding good initial image pair ============================================================================== => No good initial image pair found. Elapsed time: 1.084 [minutes] ERROR: failed to create sparse model FATAL: command failed 19:00:35 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3060 Laptop GPU [86] 19:00:35 ERROR File 'C:\Users\mylaptop\Desktop\Video\\transforms.json' does not exist. 19:00:35 WARNING Vulkan validation layer is not available. Vulkan errors will be difficult to diagnose. 19:00:36 SUCCESS Initialized Vulkan and NGX on GPU #0: NVIDIA GeForce RTX 3060 Laptop GPU
This means you did not have sufficient imagery to create a NeRF. I suggest reading this NVIDIA document on getting the proper images for NeRF creation: github.com/NVlabs/instant-ngp/blob/master/docs/nerf_dataset_tips.md
@@EveryPoint managed to fix it, it was a problem with the conda env basically my packages like numpy were not recognized outside of it which could be a path problem so i didnt use the bat files i just codes the commands my self for the training in my conda env and it worked just fine, thank you so much for the reply
Wish there was a compiled version like this of the NERF codes that takes just a few images as input (PixelNERF, DietNERF etc). Even just a tutorial for those would be very useful.
i'm wondering the same. All the papers (zip nerf etc) we can see are just research but how about the training model? it's not open source? i'm not programmer so i just dont get it
sir, i have installed as per your suggested programs but when i going to extract images from mp4 as you did in your video after first question regarding FPS they directly open the Nerf but could not able to extract images and transform jason file.
I have installed everything correctly, but I keep getting the following error message when trying to convert the video: 16:39:37 ERROR File 'C:\Users\cem_b\Desktop\Opnames' does not exist. 16:39:37 ERROR File 'Genk\VIDEO\bomen' does not exist. 16:39:37 ERROR File 'heuvel' does not exist. 16:39:37 ERROR File 'zon\transforms.json' does not exist. I've tried it with different file formats and different videos and image sequences, but I keep getting the same error. Can anyone help me out?
thanks for this! Folder of photos work great but when dragging a video file onto the nerf_video.bat i'm getting this ERROR File 'C:\Users\username\Desktop\footage\\transforms.json' does not exist I noticed you have additional folders in your instandNGP-binary folder > colmap_text and colmap_sparse, are these required for the script to work? thanks!
So I had the same issue, if you go into the instant-ngp folder, then into "scripts" there will be 2 .bat files, "download colmap" and "download ffmpeg". I believe you will need to run these before you can generate your own nerfs from videos.
Correct, however, they should automatically run when you kick off the colmap2nerf Python script. I may make a few additional to my files to check for dependencies and kick off these scripts if the requirements are not met.
@@EveryPoint Yes I have the same issue, and even when I download ffmpeg and colmap from the two .bat files it does'nt work, the transform.json file is still missing...
Coming back to this, there is a chance that colmap did not successfully process the data. This would cause colmap to run but not output a transforms file.
Thank you very much. You are the best. You are an angel from the Sky.
Рік тому+1
im getting this error even if i follow the tutorial 23:07:53 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3060 [86] 23:07:53 ERROR File 'C:\Users\dacia\Documents\New' does not exist. 23:07:53 ERROR File 'folder\New' does not exist. 23:07:53 ERROR File 'folder\transforms.json' does not exist. 23:07:53 WARNING Vulkan validation layer is not available. Vulkan errors will be difficult to diagnose. 23:07:54 SUCCESS Initialized Vulkan and NGX on GPU #0: NVIDIA GeForce RTX 3060
Okay, so it’s closing at the ffmpeg stage. I am assuming you have Python added to your path. It’s most likely either Python is not added to your path or ffmpeg is not added to your path. Can you try opening command line at your instant NGP folder and try: python scripts/colmap2nerf.py --video_in --video_fps 2 --run_colmap --aabb_scale 16
@@EveryPoint Thank you Jonathan! Note that I had to extend my path to include my anaconda terminal paths to make it run on my computer i.e. C:\Users\User_name\anaconda3; C:\Users\User_name\anaconda3\Library\mingw-w64\bin; C:\Users\User_name\anaconda3\Library\usr\bin; C:\Users\User_name\anaconda3\Library\bin; C:\Users\User_name\anaconda3\Scripts; C:\Users\User_name\anaconda3\bin;
I've waited for almost 8 days setting up the Laptop and PC into Windows + Ubuntu Development Environments to give this process as best a chance as I can. I'm still a few days away but saved today to DL the files and watch the video again to get fully excited to see the results and begin the process of what in scenes to extrapolate into the 3D render. So many filters to code in order to run a scene through different machine models to extract a single run into multiple asset needs.. Anyone that makes any of this easier is worth tossing space candy to~
I am getting an ERROR Uncaught exception: D:/a/instant-ngp/instant-ngp/dependencies/tiny-cuda-nn/include\tiny-cuda-nn/cutlass_matmul.h:332 status failed with error Error Internal. Any idea what that could be about? I'm not a programmer, so I don't know if I am supposed to install anything else? I have a GTX 1080 and a GTX 1060, Windows 10, Open GL version 4.6.0, NVIDIA 528.49, and installed CUDA Toolkit 11.6.
Sadly I couldnt get the image processing to work. Folder of images goes in, but the batch file instantly opens and closes with no Json generated. Maybe there are additional requirements for software running in the background some of us dont have by default?
Thank you for your tutorials. Question: when I use the fox dataset, no problems but when I use my dataset I get almost immediately: WARNING: No images with matches found in the database. ERROR: failed to create sparse model FATAL: command failed Why is that? Thanks
got this error Enter FPS value: 30 Traceback (most recent call last): File "C:\Users\jerem\Desktop\instant-ngp-master\\scripts\colmap2nerf.py", line 16, in import numpy as np ModuleNotFoundError: No module named 'numpy' 'C:\Users\jerem\Desktop\instant-ngp-master\\instant-ngp.exe' is not recognized as an internal or external command, operable program or batch file. C:\Users\jerem\Desktop>
Not working for me after installing all dependencies .. even importing the fox into ngp crash instant ngp . got an rtx 3060 with 6mo ram and following all requirements.
@EveryPoint Thank you so much for making these scripts! As I'm not getting any processes starting when dragging a photo folder onto the batch file (I've installed CUDA Toolkit and Vulcan, am running a 3070), I assume more is needed than the steps you outline? Installing Python and ffmpeg? Then adjusting the scripts so it all links up and executes?
I think I take for granted I have a bunch of dependencies installed. Yes, you’ll need ffmpeg and Python added to path. I’ll update the GitHub repo. I also will update the scripts to help output debug logs if it fails.
When I drop a folder of images onto your nerf_photos.bat I get an error that transform.json does not exist. I thought the batch file was supposed to create this file? Thanks kindly
Also, I’d you haven’t installed COLMAP and added it to path, there is a Python script in the instant-NGP binary scripts folder that will download and install it for you.
@@EveryPoint spaces in folder name was the problem! my goodness it works! thank you so much 💓🎉🎄 merry Christmas and happy New Year the future looks so bright now with this technology 😍 thank u for making it available for dummy like me 😜
07:44:48 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3060 Ti [86] 07:44:48 INFO Loading NeRF dataset from 07:44:48 INFO C:\Users\pc\Downloads\transforms.json 07:44:48 ERROR Uncaught exception: No training images were found for NeRF training! C:\Users\pc\Downloads> facing this error
first off thx so much for putting this together. i had a problem where when i drag a video into the video.bat, i get a prompt saying that "python was not found, run without arguments to install from microsoft store or disable this shortcut from settings". but i followed your instructions and installed python. weird thing is in anaconda when i use "python -m pip install -r requirements.txt" and it installs everything i still get the same error about python not being found. any ideas where i went wrong? thx.
Most likely one of two issues: you need to add Python to path, or your system is looking for Python3 instead of Python. I highly doubt the second one though. You can open a command prompt and type “Python -version” to see if it’s detecting Python in your path. If not? Search UA-cam on how to add Python into your path.
@@roswellcho7650 My best guess is trying installing it using: "python3 -m pip install numpy". It may be installed but not added as an environment variable.
Dumb question but where are NGP Binaries? I only see the Nvidia 3000-series (my card) download, or the NGP Master. And an error keeps repeating: import numpy as np ModuleNotFoundError: No module named 'numpy'
love this work...had a few problems, and still working around it" python -m pip install numpy i found that over here and used that python -m pip install -r requirements.txt .. found that at Instant_ngp site (linked) added pyquaternion to it
Hey EveryPoint, thanks for helping make Instant NGP even more readily available than the binaries themselves!~ I'm trying to run Instant NGP as a binary rather than in a virtual python installation. I'm running into the following error, and I was wondering if you could give me insight on how I could install numpy directly into my Windows installation... or is that not how it works (I'm really new to python, cuda, minicuda, pytorch, etc, and I always seem to run into errors ;(
Use: "python -m pip install numpy" However, you may need to install additional requirements. I suggest downloading the requirements.txt from the official Instant-NGP github repo and and installing the list. That can be done with "python -m pip install -r requirements.txt"
@@EveryPoint Thanks, EveryPoint. I managed to get it to work on my own shortly after posting the comment and a quick UA-cam/Google search here and there. I had to learn how to install numpy, colmap, etc using the cmd prompt. Additionally, I had to learn that spaces are no-good when it comes to folder trees. Currently, I'm rendering out a "scene", but all I really want is the object at the centre of my scene. I have a green screen/floor that the object is sitting on, and it's wasting time (in my opinion) trying to triangulate the points for the uniform color green background, haha... Is there a way for me to "tell" Instant NGP that I only want it to train on the object in the centre? Does the NeRF training option "Near distance" have any control over this, for example? 🤔
@@EveryPoint I was thinking if I had the coding version installed and running I just download your batch file and runs, oh boy I was wrong :D Requiements.txt done, anaconda says numpy is already exists. But still no colmap because numpy :D just the nerf window and an error that says there is no transforms.json At this point it is easier to write those two-tree lines of code to start it and fut the files in the folder :D :D :D This is still not a looser friendly solution :D :D Or maybe I should not try it after night shift :D
got this error after running few photos: ERROR Uncaught exception: FullyFusedMLP: insufficient shared memory available on the GPU. Reduce `n_neurons` or use `CutlassMLP` (better compatibility but slower) instead. need some help, any idea ? thanks
i reduced the size of all my picture from 4k to HD then changed FullyFusedMLP with CutlassMLP in the configs/ image/base.json file and it didn't changed anything to the problem.
Hello, thanks for your video. I have a problem with my camera, when I add one it goes wild and set up another frame from one I wanted, when I zoom out, I see the center of my world is not center with my NeRF, any idea how to fix this? Thanks a lot
Can this be done but with the Neurolangelo git? Looks superuseful, thanks for the script! Btw, how to get this InstantNerf into 3D mesh like OBJ, FBX, PLY, GLTF? :D
Hey thanks again for this great video. I recently set up all i need for instant NeRF and it all works finde and i already rendered stuff out and it looks good. But i also installed Vulkan SDK but when i ran instan ngp, the option for DLSS is greyed out and there is a message that Vulkan is mising at compilation. What i have to do, to make it active? Thanks for your great content. subbed. BtW, i do have a 4070ti, which has DLSS feature.
Python 15:11:18 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 4090 [89] 15:11:18 ERROR File 'C:\Users\topalov\Desktop\Instant-NGP-for-RTX-3000-and-4000\data erf\transforms.json' does not exist. 15:11:18 WARNING Vulkan: Layer name GalaxyOverlayVkLayer does not conform to naming standard (Policy #LLP_LAYER_3) 15:11:18 WARNING Vulkan: Layer name GalaxyOverlayVkLayer_VERBOSE does not conform to naming standard (Policy #LLP_LAYER_3) 15:11:18 WARNING Vulkan: Layer name GalaxyOverlayVkLayer_DEBUG does not conform to naming standard (Policy #LLP_LAYER_3
Can someone provide some help here. Everything runs (no errors) until it ends with the GUI in a black screen and an error that says that the training footage can't be loaded/allocated. The images from my input video have been generated in the folder (505 frames in 4k). Also, loading the generated ".json" file into a newly started ngp GUI gives the same result. Really curious what's going wrong here. I'm running the ngp rtx2000 version for my rtx2060. Maybe I should down-res my footage to 2k!?
I see what's the issue. I wasted hours trying to do this with 360 photos, and it wouldn't create the transforms.json. Then I tried with normal photos and it does. So, yeah, it doesn't work for 360... Anybody has any idea how to do this with 360? I mean even nvidia features it on the repo (flythroughs done with 360 photos)
I'm very good at creating user interfaces (with 10 years of experience) maybe you want cooperate and create proper interface for it so we everyone can use it ? :)
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=
Hi, I just couldn't get it to work. I was constantly running into errors while executing colmap. However, I found a workaround, and I wanted to share it (maybe it helps someone who is also struggling) ua-cam.com/video/l5q1uo1A368/v-deo.html But of course your video would be the more proper and direct way :-) Greetings!!!
Sorry but I did about 6 tests of instant-ngp using the photo sets I took with my Nikon DSL and used in 3DFZephyr to create obj textured meshs. The results I got from 3DFZephyr where a 100 times superior to anything in instant-ngp. Not sure what all this hype about nerfs are. The results are inferior and it is a pain to install and setup instant-ngp compared to 3DFZephyr, Metashape, Reality Capture or Meshroom. Nefs IMHO are all smoke and mirrors and not worth the effort.
Brilliant Vid making the barrier of entry easier for all. Quick tech question though I seem to be getting a weird error. Would it be possible to have a look at it? running colmap with: db=colmap.db images="C:\Users\Slave-0\Desktop\AI programs\Instant NGP NeRF\data\firsttest\Images" sparse=colmap_sparse text=colmap_text warning! folders 'colmap_sparse' and 'colmap_text' will be deleted/replaced. continue? (Y/n)y ==== running: colmap feature_extractor --ImageReader.camera_model OPENCV --ImageReader.camera_params "" --SiftExtraction.estimate_affine_shape=true --SiftExtraction.domain_size_pooling=true --ImageReader.single_camera 1 --database_path colmap.db --image_path "C:\Users\Slave-0\Desktop\AI programs\Instant NGP NeRF\data\firsttest\Images" programs\Instant was unexpected at this time. FATAL: command failed 'C:\Users\Slave-0\Desktop\AI' is not recognized as an internal or external command, operable program or batch file.
Thank you for the great tutorial. I got an error saying: Uncaught exception: D:/a/instant-ngp/instant-ngp/dependencies/tiny-cuda-nn/include\tiny-cuda-nn/cutlass_matmul.h:332 status failed with error Error Internal Do yo have any idea why is that happening? thank you!
When I hit 2 on the video drag drop nerf_video.bat. I get this: 20:32:57 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3090 [86] 20:32:57 ERROR File 'C:\Users\Admin\OneDrive\Desktop\desktopttest\transforms.json' does not exist. 20:32:57 WARNING Vulkan instance validation layer is not available. Vulkan errors will be difficult to diagnose. 20:32:57 SUCCESS Initialized Vulkan and NGX on GPU #0: NVIDIA GeForce RTX 3090 20:32:57 SUCCESS Initialized OpenGL version 4.6.0 NVIDIA 531.14 NGP opens blank screen. Running with nothing. Nor did it generate images to use from said video. A comment below said to add python from the Microsoft store to fix it but it didn't fix it. What am I missing?
Amazing video! But I keep getting this strange error. 'C:\Users\steve\OneDrive\Desktop\InstantNGPforRTX2000' is not recognized as an internal or external command, operable program or batch file. FATAL: command failed 'C:\Users\steve\OneDrive\Desktop\InstantNGPforRTX2000' is not recognized as an internal or external command, operable program or batch file.
Thanks you for the requirement bat! , I didn't noticed the requirement bat at first and I was shocked by so much tech talk so I have been all day installing dependencies one by one (as a 3D artist I didn't even had anaconda installed), don't be like me guys, just put the requirements bat file in the instant ngp folder and say thanks hahahaha
Thanks! I really with the video actually said something about this...
ERROR File 'H: \transforms.json' does not exist.
They forgot to mention in the video that the folder containing the images NEEDS to be named "images". Hope this helps!
Thank you very much! This problem has been bothering me for a long time, but because I don't know how to program, I haven't been able to make the program run properly. I really appreciate your help!
Small note from a noob here. Don't put spaces in your folder names. I learned that the fun way
That has been added to the Readme instructions
Awesome!!! Thank you so much for making both the video and the .bat files
You're welcome!
Seeing this makes me wonder if you could run an instance of instant NGP in a cloud GPU provider like Lambda Labs. That way, you could still play use this tool even if you don't have a compatible setup. Export all necessary files to mess with in Blender or something. Not sure how much functionality might be lost though. Would be interested to hear your thoughts on this.
If I might suggest next time you make a video, plz increase the font size or something. Some of us have bad eyes.
Great video btw.
Definitely. The written instructions are in the GitHub repo too. The text was way too small.
Great follow up video! I noticed that the windows zip Nvidia provided didn't include the required scripts for downloading COLMAP if you didn't already have it installed and tried to leave a comment about how I got around it but it seems like it got blocked. Glad you put this together to help simplify the process!
Thanks for the update, I’ll include instructions later today on how to download Colmap and install it. It’s fairly easy.
@@EveryPoint Looks like they fixed the exclusion of the "download COLMAP" scripts being missing from the binaries two days ago, so I guess that's a non-issue now. Now my struggle is to convince colmap to use CUDA rather than the CPU. Guess its time to go back an RTFM. :)
@@msampson3d Or the old WADC trick, if nothing else helps...
@@msampson3d You can also capture data using the Record-3D app and skip the whole COLMAP sequence.
I was kinda waiting for something like this to dive in, thanks a lot, jumping in tomorrow !!
C:\Users\******\Desktop
erfik
oom: Permission denied
FATAL: command failed
18:54:21 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3070 Ti [86]
18:54:21 ERROR File 'C:\Users\********\Desktop
erfik\\transforms.json' does not exist.
18:54:22 SUCCESS Initialized Vulkan and NGX on GPU #0: NVIDIA GeForce RTX 3070 Ti
are u solve？
same problem :/
I need help how to fix this
Finding good initial image pair
==============================================================================
=> No good initial image pair found.
Elapsed time: 1.084 [minutes]
ERROR: failed to create sparse model
FATAL: command failed
19:00:35 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3060 Laptop GPU [86]
19:00:35 ERROR File 'C:\Users\mylaptop\Desktop\Video\\transforms.json' does not exist.
19:00:35 WARNING Vulkan validation layer is not available. Vulkan errors will be difficult to diagnose.
19:00:36 SUCCESS Initialized Vulkan and NGX on GPU #0: NVIDIA GeForce RTX 3060 Laptop GPU
did u manage to fix this?
This means you did not have sufficient imagery to create a NeRF. I suggest reading this NVIDIA document on getting the proper images for NeRF creation: github.com/NVlabs/instant-ngp/blob/master/docs/nerf_dataset_tips.md
transforms.json' does not exist i get this error when using the bat files any idea on how to fix? the fox images work just fine
Most likely COLMAP did not run successfully. Did you have quite a few overlapping images or good parallax motion in your video?
@@EveryPoint managed to fix it, it was a problem with the conda env basically my packages like numpy were not recognized outside of it which could be a path problem so i didnt use the bat files i just codes the commands my self for the training in my conda env and it worked just fine, thank you so much for the reply
@@sashy572 can you tell me how you did it tho? im still super new to this and scratching my head trying to solve this issue :)
Wish there was a compiled version like this of the NERF codes that takes just a few images as input (PixelNERF, DietNERF etc). Even just a tutorial for those would be very useful.
i'm wondering the same. All the papers (zip nerf etc) we can see are just research but how about the training model? it's not open source? i'm not programmer so i just dont get it
sir, i have installed as per your suggested programs but when i going to extract images from mp4 as you did in your video after first question regarding FPS they directly open the Nerf but could not able to extract images and transform jason file.
Man.... I love you!
thxxxxx
I have installed everything correctly, but I keep getting the following error message when trying to convert the video:
16:39:37 ERROR File 'C:\Users\cem_b\Desktop\Opnames' does not exist.
16:39:37 ERROR File 'Genk\VIDEO\bomen' does not exist.
16:39:37 ERROR File 'heuvel' does not exist.
16:39:37 ERROR File 'zon\transforms.json' does not exist.
I've tried it with different file formats and different videos and image sequences, but I keep getting the same error. Can anyone help me out?
i got error no transform.json
me too
@@sulaimanaldokhayel same what does that mean
thanks for this! Folder of photos work great but when dragging a video file onto the nerf_video.bat i'm getting this
ERROR File 'C:\Users\username\Desktop\footage\\transforms.json' does not exist
I noticed you have additional folders in your instandNGP-binary folder > colmap_text and colmap_sparse, are these required for the script to work? thanks!
So I had the same issue, if you go into the instant-ngp folder, then into "scripts" there will be 2 .bat files, "download colmap" and "download ffmpeg". I believe you will need to run these before you can generate your own nerfs from videos.
Correct, however, they should automatically run when you kick off the colmap2nerf Python script. I may make a few additional to my files to check for dependencies and kick off these scripts if the requirements are not met.
@@EveryPoint Yes I have the same issue, and even when I download ffmpeg and colmap from the two .bat files it does'nt work, the transform.json file is still missing...
@@sreameight me too
Coming back to this, there is a chance that colmap did not successfully process the data. This would cause colmap to run but not output a transforms file.
I'll definitely try this out thanks!
Thank you very much. You are the best. You are an angel from the Sky.
im getting this error even if i follow the tutorial 23:07:53 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3060 [86]
23:07:53 ERROR File 'C:\Users\dacia\Documents\New' does not exist.
23:07:53 ERROR File 'folder\New' does not exist.
23:07:53 ERROR File 'folder\transforms.json' does not exist.
23:07:53 WARNING Vulkan validation layer is not available. Vulkan errors will be difficult to diagnose.
23:07:54 SUCCESS Initialized Vulkan and NGX on GPU #0: NVIDIA GeForce RTX 3060
Do you have a space in a folder name?
Are you solve？
Il try to change system language to English and reinstall it. Hope it will Work
it keeps closing everytime I place my file in the video.bat file, the pictures.bat doesnt even open i have cuda downloaded
I’ll have to add a log file export for when it doesn’t run successfully. Do you have ffmpeg and Colmap pre-installed on your PC?
What is the video file type? Also, was an images folder created in the same folder as the source video?
Okay, so it’s closing at the ffmpeg stage. I am assuming you have Python added to your path. It’s most likely either Python is not added to your path or ffmpeg is not added to your path.
Can you try opening command line at your instant NGP folder and try: python scripts/colmap2nerf.py --video_in --video_fps 2 --run_colmap --aabb_scale 16
@@EveryPoint how would I add python to the path? I have no coding skills or much understanding of how this stuff works
@@EveryPoint Thank you Jonathan! Note that I had to extend my path to include my anaconda terminal paths to make it run on my computer i.e.
C:\Users\User_name\anaconda3;
C:\Users\User_name\anaconda3\Library\mingw-w64\bin;
C:\Users\User_name\anaconda3\Library\usr\bin;
C:\Users\User_name\anaconda3\Library\bin;
C:\Users\User_name\anaconda3\Scripts;
C:\Users\User_name\anaconda3\bin;
I've waited for almost 8 days setting up the Laptop and PC into Windows + Ubuntu Development Environments to give this process as best a chance as I can. I'm still a few days away but saved today to DL the files and watch the video again to get fully excited to see the results and begin the process of what in scenes to extrapolate into the 3D render. So many filters to code in order to run a scene through different machine models to extract a single run into multiple asset needs.. Anyone that makes any of this easier is worth tossing space candy to~
Awesome!!! Thank you so much for sharing this video
You're welcome!
im lost, its keep shutdown when i drag fox folder to my instant ngp. can you do more clear tutorial from scratch
same
I am getting an ERROR Uncaught exception: D:/a/instant-ngp/instant-ngp/dependencies/tiny-cuda-nn/include\tiny-cuda-nn/cutlass_matmul.h:332 status failed with error Error Internal. Any idea what that could be about? I'm not a programmer, so I don't know if I am supposed to install anything else? I have a GTX 1080 and a GTX 1060, Windows 10, Open GL version 4.6.0, NVIDIA 528.49, and installed CUDA Toolkit 11.6.
Sadly I couldnt get the image processing to work. Folder of images goes in, but the batch file instantly opens and closes with no Json generated. Maybe there are additional requirements for software running in the background some of us dont have by default?
Did a video work? Also, did you have any spaces in the folder or sub folder names?
@@EveryPoint Your efforts are amazing, the work you are doing is top class. Keep it up mate !!
@@matthewmooney1935 Thank you!
@@EveryPoint Thank you for this ! I solved the problem by changing user file name (be carefull if you see this message and want to do it)
Can you please explain what is meant by «adding Colemap and FFMPEG to path»?
Thanks so much for the script, very useful
Thank you for your tutorials.
Question: when I use the fox dataset, no problems but when I use my dataset I get almost immediately:
WARNING: No images with matches found in the database.
ERROR: failed to create sparse model
FATAL: command failed
Why is that?
Thanks
Thank you so much! Here goes my sleep 😂
Okay, this only takes minutes to get up and running. I swear it's that easy. Just sit back and let it do its magic.
got this error Enter FPS value: 30
Traceback (most recent call last):
File "C:\Users\jerem\Desktop\instant-ngp-master\\scripts\colmap2nerf.py", line 16, in
import numpy as np
ModuleNotFoundError: No module named 'numpy'
'C:\Users\jerem\Desktop\instant-ngp-master\\instant-ngp.exe' is not recognized as an internal or external command,
operable program or batch file.
C:\Users\jerem\Desktop>
do pip install numpy
Not working for me after installing all dependencies .. even importing the fox into ngp crash instant ngp . got an rtx 3060 with 6mo ram
and following all requirements.
@EveryPoint Thank you so much for making these scripts! As I'm not getting any processes starting when dragging a photo folder onto the batch file (I've installed CUDA Toolkit and Vulcan, am running a 3070), I assume more is needed than the steps you outline? Installing Python and ffmpeg? Then adjusting the scripts so it all links up and executes?
I think I take for granted I have a bunch of dependencies installed. Yes, you’ll need ffmpeg and Python added to path. I’ll update the GitHub repo. I also will update the scripts to help output debug logs if it fails.
@@EveryPoint Fantastic, thank you so much!
sorry for the cuestion, I'm simply a 3d artist, very new to this, how i add the path (that you said in the readme) correctly to try this??
When I drop a folder of images onto your nerf_photos.bat I get an error that transform.json does not exist. I thought the batch file was supposed to create this file?
Thanks kindly
After instally numpy (pip install numpy) and opencv (pip install opencv-pyjthon) the nerf_photos.bat is working.
it keeps closing after i am trying to drag video, doesn't ask to install colmap
My chance there is a space in the folder path? This may be the issue.
Also, I’d you haven’t installed COLMAP and added it to path, there is a Python script in the instant-NGP binary scripts folder that will download and install it for you.
@@EveryPoint spaces in folder name was the problem!
my goodness it works!
thank you so much 💓🎉🎄 merry Christmas and happy New Year the future looks so bright now with this technology 😍 thank u for making it available for dummy like me 😜
07:44:48 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3060 Ti [86]
07:44:48 INFO Loading NeRF dataset from
07:44:48 INFO C:\Users\pc\Downloads\transforms.json
07:44:48 ERROR Uncaught exception: No training images were found for NeRF training!
C:\Users\pc\Downloads>
facing this error
ERROR File 'H: \transforms.json' does not exist.
@@isaacarroyoleon103 same as mine
first off thx so much for putting this together. i had a problem where when i drag a video into the video.bat, i get a prompt saying that "python was not found, run without arguments to install from microsoft store or disable this shortcut from settings". but i followed your instructions and installed python.
weird thing is in anaconda when i use "python -m pip install -r requirements.txt" and it installs everything i still get the same error about python not being found. any ideas where i went wrong? thx.
Most likely one of two issues: you need to add Python to path, or your system is looking for Python3 instead of Python. I highly doubt the second one though. You can open a command prompt and type “Python -version” to see if it’s detecting Python in your path. If not? Search UA-cam on how to add Python into your path.
@@roswellcho7650 My best guess is trying installing it using: "python3 -m pip install numpy". It may be installed but not added as an environment variable.
Hey there, so I guess since we download the latest version from Nvidia's side, it's running the latest version, right?
Dumb question but where are NGP Binaries? I only see the Nvidia 3000-series (my card) download, or the NGP Master. And an error keeps repeating: import numpy as np
ModuleNotFoundError: No module named 'numpy'
love this work...had a few problems, and still working around it"
python -m pip install numpy i found that over here and used that
python -m pip install -r requirements.txt .. found that at Instant_ngp site (linked)
added pyquaternion to it
I added a batch script to address this.
Great job! Adding it to my work/project queue… Thanks! PS, didn’t know you lived so close (only 45 minutes south of me).
PS only 30 minutes north as you as well.
Thanks! We’re so close!
@@EveryPoint You guys in SoCal?
@@andershjemdahl6547 We're everywhere. :P Actually, we are mainly distributed in the Pacific Northwest, Atlanta, and North Carolina.
Hey EveryPoint, thanks for helping make Instant NGP even more readily available than the binaries themselves!~
I'm trying to run Instant NGP as a binary rather than in a virtual python installation.
I'm running into the following error, and I was wondering if you could give me insight on how I could install numpy directly into my Windows installation... or is that not how it works (I'm really new to python, cuda, minicuda, pytorch, etc, and I always seem to run into errors ;(
Use: "python -m pip install numpy"
However, you may need to install additional requirements. I suggest downloading the requirements.txt from the official Instant-NGP github repo and and installing the list. That can be done with "python -m pip install -r requirements.txt"
@@EveryPoint Thanks, EveryPoint. I managed to get it to work on my own shortly after posting the comment and a quick UA-cam/Google search here and there. I had to learn how to install numpy, colmap, etc using the cmd prompt. Additionally, I had to learn that spaces are no-good when it comes to folder trees.
Currently, I'm rendering out a "scene", but all I really want is the object at the centre of my scene. I have a green screen/floor that the object is sitting on, and it's wasting time (in my opinion) trying to triangulate the points for the uniform color green background, haha...
Is there a way for me to "tell" Instant NGP that I only want it to train on the object in the centre? Does the NeRF training option "Near distance" have any control over this, for example?
🤔
@@EveryPoint I was thinking if I had the coding version installed and running I just download your batch file and runs, oh boy I was wrong :D Requiements.txt done, anaconda says numpy is already exists. But still no colmap because numpy :D just the nerf window and an error that says there is no transforms.json At this point it is easier to write those two-tree lines of code to start it and fut the files in the folder :D :D :D This is still not a looser friendly solution :D :D Or maybe I should not try it after night shift :D
@@mihalydozsa2254 did u manage to find a fix for this? i have the same problem
@@sashy572 no I gave up after my comment. (at this point the old command line version would not work either :D)
got this error after running few photos:
ERROR Uncaught exception: FullyFusedMLP: insufficient shared memory available on the GPU. Reduce `n_neurons` or use `CutlassMLP` (better compatibility but slower) instead.
need some help, any idea ? thanks
i reduced the size of all my picture from 4k to HD then changed FullyFusedMLP with CutlassMLP in the configs/ image/base.json file and it didn't changed anything to the problem.
I can't get colmap to install. There's a url that just doesn't seem to exist and it just won't download all the files required.
Could not find '.json' file.
Where is the problem please?
running ffmpeg with input video file="C:
gp\instant-ngp\data\su\su.mp4", output image folder="C:
gp\instant-ngp\data\su\images", fps=2.0.
warning! folder '"C:
gp\instant-ngp\data\su\images"' will be deleted/replaced. continue? (Y/n)
==== running: mkdir "C:
gp\instant-ngp\data\su\images"
==== running: ffmpeg -i "C:
gp\instant-ngp\data\su\su.mp4" -qscale:v 1 -qmin 1 -vf "fps=2.0" "C:
gp\instant-ngp\data\su\images"/%04d.jpg
FATAL: command failed
16:19:32 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3090 [86]
16:19:32 ERROR File 'C:
gp\instant-ngp\data\su\transforms.json' does not exist.
16:19:32 SUCCESS Initialized OpenGL version 4.6.0 NVIDIA 531.14
did you find the solution to the error .json doesn't exist?
Awesome bro, you rock!
Subbed, keep up.
Hello, thanks for your video. I have a problem with my camera, when I add one it goes wild and set up another frame from one I wanted, when I zoom out, I see the center of my world is not center with my NeRF, any idea how to fix this?
Thanks a lot
Sweet enjoying the sessions
Glad you found it informative!
Can this be done but with the Neurolangelo git? Looks superuseful, thanks for the script! Btw, how to get this InstantNerf into 3D mesh like OBJ, FBX, PLY, GLTF? :D
thank you! SUBSCRIBED
Thank you! That means so much to us!
thank you thank you !!! you're a genius!!! i wo't sleeo tonight!
Have fun!
Thanks!!!! Por fin pude usar instant ngp!
Hey thanks again for this great video. I recently set up all i need for instant NeRF and it all works finde and i already rendered stuff out and it looks good. But i also installed Vulkan SDK but when i ran instan ngp, the option for DLSS is greyed out and there is a message that Vulkan is mising at compilation. What i have to do, to make it active? Thanks for your great content. subbed. BtW, i do have a 4070ti, which has DLSS feature.
Python 15:11:18 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 4090 [89]
15:11:18 ERROR File 'C:\Users\topalov\Desktop\Instant-NGP-for-RTX-3000-and-4000\data
erf\transforms.json' does not exist.
15:11:18 WARNING Vulkan: Layer name GalaxyOverlayVkLayer does not conform to naming standard (Policy #LLP_LAYER_3)
15:11:18 WARNING Vulkan: Layer name GalaxyOverlayVkLayer_VERBOSE does not conform to naming standard (Policy #LLP_LAYER_3)
15:11:18 WARNING Vulkan: Layer name GalaxyOverlayVkLayer_DEBUG does not conform to naming standard (Policy #LLP_LAYER_3
are u solve?
@@GAMING_DEV-wr6rp ah you saved me!
You're great!!! Thanks
Can someone provide some help here.
Everything runs (no errors) until it ends with the GUI in a black screen and an error that says that the training footage can't be loaded/allocated. The images from my input video have been generated in the folder (505 frames in 4k). Also, loading the generated ".json" file into a newly started ngp GUI gives the same result. Really curious what's going wrong here. I'm running the ngp rtx2000 version for my rtx2060.
Maybe I should down-res my footage to 2k!?
I see what's the issue. I wasted hours trying to do this with 360 photos, and it wouldn't create the transforms.json. Then I tried with normal photos and it does. So, yeah, it doesn't work for 360...
Anybody has any idea how to do this with 360? I mean even nvidia features it on the repo (flythroughs done with 360 photos)
#Gold Big 'THANK YOU' 💯👍
😃
Thanks a lot!
can you help me?
I'm very good at creating user interfaces (with 10 years of experience) maybe you want cooperate and create proper interface for it so we everyone can use it ? :)
gtx 1650 cannot afford... 😞
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=
This make me more confused 😅
А как установить то всю эту папку с COLMAP?бесполезное видео
*promosm* 🌟
Hi, I just couldn't get it to work. I was constantly running into errors while executing colmap.
However, I found a workaround, and I wanted to share it (maybe it helps someone who is also struggling)
ua-cam.com/video/l5q1uo1A368/v-deo.html
But of course your video would be the more proper and direct way :-)
Greetings!!!
Sorry but I did about 6 tests of instant-ngp using the photo sets I took with my Nikon DSL and used in 3DFZephyr to create obj textured meshs. The results I got from 3DFZephyr where a 100 times superior to anything in instant-ngp. Not sure what all this hype about nerfs are. The results are inferior and it is a pain to install and setup instant-ngp compared to 3DFZephyr, Metashape, Reality Capture or Meshroom. Nefs IMHO are all smoke and mirrors and not worth the effort.
Brilliant Vid making the barrier of entry easier for all.
Quick tech question though I seem to be getting a weird error. Would it be possible to have a look at it?
running colmap with:
db=colmap.db
images="C:\Users\Slave-0\Desktop\AI programs\Instant NGP NeRF\data\firsttest\Images"
sparse=colmap_sparse
text=colmap_text
warning! folders 'colmap_sparse' and 'colmap_text' will be deleted/replaced. continue? (Y/n)y
==== running: colmap feature_extractor --ImageReader.camera_model OPENCV --ImageReader.camera_params "" --SiftExtraction.estimate_affine_shape=true --SiftExtraction.domain_size_pooling=true --ImageReader.single_camera 1 --database_path colmap.db --image_path "C:\Users\Slave-0\Desktop\AI programs\Instant NGP NeRF\data\firsttest\Images"
programs\Instant was unexpected at this time.
FATAL: command failed
'C:\Users\Slave-0\Desktop\AI' is not recognized as an internal or external command,
operable program or batch file.
Try removing spaces from your folder names.
@@EveryPoint Yup, that seems to have done it. Such a simple fix XD
Thank you for the great tutorial. I got an error saying: Uncaught exception: D:/a/instant-ngp/instant-ngp/dependencies/tiny-cuda-nn/include\tiny-cuda-nn/cutlass_matmul.h:332 status failed with error Error Internal
Do yo have any idea why is that happening? thank you!
When I hit 2 on the video drag drop nerf_video.bat. I get this:
20:32:57 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA GeForce RTX 3090 [86]
20:32:57 ERROR File 'C:\Users\Admin\OneDrive\Desktop\desktopttest\transforms.json' does not exist.
20:32:57 WARNING Vulkan instance validation layer is not available. Vulkan errors will be difficult to diagnose.
20:32:57 SUCCESS Initialized Vulkan and NGX on GPU #0: NVIDIA GeForce RTX 3090
20:32:57 SUCCESS Initialized OpenGL version 4.6.0 NVIDIA 531.14
NGP opens blank screen. Running with nothing. Nor did it generate images to use from said video.
A comment below said to add python from the Microsoft store to fix it but it didn't fix it. What am I missing?
Amazing video! But I keep getting this strange error.
'C:\Users\steve\OneDrive\Desktop\InstantNGPforRTX2000' is not recognized as an internal or external command,
operable program or batch file.
FATAL: command failed
'C:\Users\steve\OneDrive\Desktop\InstantNGPforRTX2000' is not recognized as an internal or external command,
operable program or batch file.
Do you have a space in any of your folder names?
@@EveryPoint I do not, but i seem to have fixed it by removing the parentheses and other special characters in the folder names
@@i_hatecheese8665 To clarify, you removed the quotation marks from the batch scripts? I may need to clean up the syntax.
@@EveryPoint I removed the special characters that come with the base folder. The batch scripts were fine, just the folder names were off
@@i_hatecheese8665 I added to the documentation to ensure you do not have spaces or special characters in folder and files names.