It's pretty tough to remember all the codes for Instant-NeRF so I have the ones I use compiled here. Anything bolded below represents what to edit based upon your specific file/need.
Transform Guide
When starting with a image set that you would like to create in Instant-NeRF, there are a couple of steps that can save a lot of time. The first is to take your VRAM and make sure that the number of images you are using will not exceed your memory. If the amount of images exceed your available VRAM, you will either need to lower your aabb or downsample your images. Depending on your use case, either option will be viable. To that same point, if you have a small object that you are NeRFing, if can be good to lower your aabb. Once you have all of these decided, simply include the folder in which your images are hosted after data/ and you'll be ready to go.
python scripts/colmap2nerf.py --colmap_matcher exhaustive --run_colmap --aabb_scale 16 --images data/insertfoldernamehere
Transform from a Video Code
In order to run the transform, from a video, please make sure that you have installed FFMPEG already and have it set up to run. After this is complete, drop the video you would like to use into the "Scripts" folder within Instant-NeRF. Alter the code below where it says videoname.mov with your file name and file type. It has to match exactly, so if you are using a .mp4 or another video type, make sure to specify it here. You will also need to instruct how many frames per second (fps) the script will pull. If you're having trouble coming up with how many frames you can use, check out this calculator!
python colmap2nerf.py --video_in videoname.mov --video_fps 2.5 --run_colmap --aabb_scale 16
Launch Instant-NeRF
After the transform is complete and the folder is set up, you can launch Instant-NeRF with the below code. It is similar to the Image Transform Guide in that you need to fill in your folder name at the end of data/
build\testbed.exe --scene data/yourfoldername
Rendering Guide
Once you have a camera path that you are satisfied with in Instant-NGP, you can use the following code to begin the render. Please make sure to update the foldername to where the folder of where your NeRF is enclosed and to rename the sample for whatever you want the exported video file to be called.
Protip! If you are rendering the same file multiple times, make sure to change the name everytime of your output name otherwise it will overwrite your video file. You can also customize your resolution to whatever you like, but please be aware that the higher the resolution, the longer the time it will take to render out.
In addition, the length of the video (seconds) and frames per second (fps) are customizable. However, the script natively renders it into 24fps, so if you want a 10 second video at 24 frames per second, you can leave it as is. However, if you wanted it to be at 30fps, you would put 12.5 seconds, which would result to the 10 second export you were looking for.
python scripts/render.py --scene data/foldername--n_seconds 8 --fps 24 --render_name sample --width 3840 --height 2160
Python Extensions
If you are familiar with editing the code, you can use Python 3.7 or higher to make interactive bindings to simplify your workflow such as launching Instant-NeRF with a snapshot.
