support fx and fy

[Khoa-NT]

Hi, Thank you for the amazing paper and for sharing the code. From this, I saw the code is using single focal length f. Currently, my colmap result returned fx & fy. I would like to ask if is it ok to run with only fx from your code. Or can you support the fx and fy?

[tcavallari]

Hello, thanks for your feedback.

You are correct that the code at the moment uses a single focal length for both x and y. There are two options:

• If your COLMAP focal lengths are similar enough, you could average them and use the code as it is.
• Alternatively, you could store them both in the various calibration files (separated by space), then update this line: https://github.com/nianticlabs/ace/blob/main/dataset.py#L365 to read them both as a tensor of two elements. You also need to update these two lines: https://github.com/nianticlabs/ace/blob/main/dataset.py#L487 to use them both in the camera calibration matrix. The rest of the training/testing code should work out of the box.

[Khoa-NT]

Hello @tcavallari, Thank you for your prompt reply and detailed solutions. I'm currently making a script to convert COLMAP binary to your pgt format to train the model with my customs dataset. I will try both of the solutions and update you later.