Rintarooo / pyramidify

Plot camera poses in 3D for creating NeRFs.

pyramidify

An interactive notebook for placing camera poses (pyramids) in your scene for rendering NeRFs.

Here's a guide for using NVIDIA NGP without their GUI (using a Colab GPU):

Step 1: Images

• Take some images of a scene using any camera.
• Try to revolve around an origin rather than making panoramas.
• Make sure the images aren't blurry.
• Around 30-50 images should be enough.

Step 2: COLMAP

• Run COLMAP Automatic Reconstruction on the images.
• Export your model as text files.
• You should get these:

  camera.txt
  images.txt
  points3D.txt

Step 3: transforms.json

• Place images and text files like this:

  images
    0001.jpeg
    0002.jpeg
    ...
  data
    camera.txt
    images.txt
    points3D.txt

• Run transforms.ipynb.

Step 4: base_cam.json

• Place camera poses for rendering using base_cam.ipynb.
• Change FOV accordingly.

Step 5: Nvidia NGP

• Open NGP.ipynb in Google Colab.
• Turn on GPU in runtime.
• Upload images, transforms.json and base_cam.json.
• You should get something like this: