Since an orthographic camera does not have a focal length or FOVY, how should I proceed if I want to use images rendered by an orthographic camera for training? I would greatly appreciate any insights you can provide.
Yes. Since we use ray-splat intersection, I think our method can support various camera models, even non-linear cameras. But it may need rewriting the cuda code for supporting this.
Since an orthographic camera does not have a focal length or FOVY, how should I proceed if I want to use images rendered by an orthographic camera for training? I would greatly appreciate any insights you can provide.