Dear authors, thanks for your code!
I found that the pretrained model droid.pth provided by the official performs very poorly in the ME004 scenario in tartanair, with an ate value of 12, which is far from the value mentioned in DeepVO. I have also tried to train the model myself, but it is difficult to converge to results similar to those in other scenarios.
Dear authors, thanks for your code! I found that the pretrained model droid.pth provided by the official performs very poorly in the ME004 scenario in tartanair, with an ate value of 12, which is far from the value mentioned in DeepVO. I have also tried to train the model myself, but it is difficult to converge to results similar to those in other scenarios.