Inference on a dataset with different camera parameters

[aditya18007]

Suppose I use the network trained on KITTI dataset, Can I use it to run inference on nuScenes. It appears to me that the results will be bad. Mostly because the Camera parameters Matrix is used in training the dataset. So If I change the Matrix at evaluation, will the results be good? Am I wrong on the assumption ?