Spectacular Rec for Gaussian Splatting and NeRF reconstructions (iPhone)

Hey fellow Nerf and Gaussian Splatting  enthusiastics in this video we'll go   over how to use Spectacular Rec  together with Nerfstudio first   download the app from the app store  and then we'll record a video using it you should try to move the camera slowly to  avoid motion blur the camera brightness and   color temperature settings are locked during  the recording and then Focus distance is also   locked so try to stay at the same distance  from the target to avoid image going out of focus keeping all the camera settings  locked produces more stable reconstructions once the recording is done the app allows you to   upload the recorded videos to  clad storage of your choosing if you have a lot of recordings it might  be faster to connect the phone to your PC   directly via cable and transfer them in file  manager once the files are on your computer   you need to extract them you can also inspect  that the video looks good next we'll install   the SDK examples git repository from GitHub  we'll use the replay to Nerf script to compute   the camera trajectory and Export the data  into format that the Nerf Studio understands the script has some dependencies so install  those into python virtual environment you'll get a 3D preview of the camera  directory while it's being computed   this will take some time to produce  the video you see here is sped up once the preliminary path is ready  it's further optimized this might   take couple minutes now we have  the data in the right format The   Next Step is to install the Nerf  studio if you don't have it yet this can take a while and we'll skip  it here to save time once Nerf studio   is set up we'll use the NS train to  start the training process use the   directory given by the replay to n  NF python script we used previously you can open Nerf studio in a browser  to see the training in progress here you can see the finished Nerf reconstruction   you can easily set up a camera  path and render out the video that's all from me thank you for watching and  please share your results using the pipeline shown here

Channel: Spectacular AI

Views: 2,473

Keywords:

Id: d77u-E96VVw

Length: 4min 46sec (286 seconds)

Published: Tue Dec 05 2023