Aruco2nerf script

[juanmacaaz]

First of all, I hope you are all doing great.

I have dedicated myself to using aruco to find the POSE of the cameras to train a scene in NERF and I have created a script that creates a transform file in the same standard as intant ngp.

1. We need to calibrate the camera, for this we can use the following command:

python .\aruco2nerf.py --folder_dir calibration_imgs --calib

In the folder_dir argument we have to indicate the images to calibrate the camera. We use charuco for the calibration. As you can see in the image

2. Once calibrated the camera, we can obtain the poses with the following code:

python .\aruco2nerf.py --folder_dir images

In my case I have created a 3x3 arucos board where in the center you can place the object, as you can see in the following image.

Thank you very much.

[christian-ferre]

Ty so much, a really great job ;=)

[satpalsr]

@juanmacaaz Are results better than colmap? I am getting similar ok kinda result with both.

[juanmacaaz]

@juanmacaaz Are results better than colmap? I am getting similar ok kinda result with both.

The results are similar, the main difference is that colmap needs an exponential execution time, while using aruco this time becomes linear.

[Saddy21]

@juanmacaaz Are results better than colmap? I am getting similar ok kinda result with both.

The results are similar, the main difference is that colmap needs an exponential execution time, while using aruco this time becomes linear.

Do the dimensions match the actual model dimensions using this method,as in colmap,it is up scaling the model. Thanks