Known intrinsics not working as I expected

[dsvilarkovic]

Hello, Following your README.md it says that in case I have my own ground truth intrinsics, I should define them as:

intrinsics_path = "assets/demo/intrinsics.npy"

# Load the intrinsics if available
intrinsics = torch.from_numpy(np.load(intrinsics_path)) # 3 x 3

predictions = model.infer(rgb, intrinsics)

I tried this in a small demo.ipynb, but I notice that my gt_intrinsics is corrected afterwards:

Is there any way I can depth map estimation without changing my gt-intrinsics?

[lpiccinelli-eth]

Thank you for raising this issue.

You can check issue #48 for a more detailed explanation, in short: the intrinsic returned by the model are the estimated ones, if you provide some GT intrinsics the model will use the GT, but it will return the predicted ones anyway.

[dsvilarkovic]

All right, thanks for repeating the answer! I will close this issue, and test your module in the upcoming week, hoping it helps for dense keypoint initialization for splatting!