Hello, while reading your paper again, I have a question and I hope you can have time to reply to me.I would greatly appreciate it.
I think the evaluation metrics for stereo training models will be higher than those for monocular training models.
I think the stereo training model has more information.
However, in models with certain metrics lower than monocular training.
What is the reason for this?
Hello, while reading your paper again, I have a question and I hope you can have time to reply to me.I would greatly appreciate it. I think the evaluation metrics for stereo training models will be higher than those for monocular training models. I think the stereo training model has more information. However, in models with certain metrics lower than monocular training. What is the reason for this?