Implementing NeRF from scratch, by following the original research paper.
I first started out by creating a 2D "nerf", which just memorizes a photo: https://github.com/shahanneda/NeRfFromScratch/blob/main/single-image.ipynb
And then I moved on to making it 3D and an actual nerf with volumetric rendering: https://github.com/shahanneda/NeRfFromScratch/blob/main/3d_image.ipynb
Then, to run overnight training runs, I extracted some code to classes and more proper python files, and also integrated with weights and biases to run many different training runs with different parameters:
