I tried to add my own module to BARF, but during the implementation, I found that the model converges normally during training, but the PSNR is very low during validation. Through camera visualization, I can see that the camera poses are all well learned during training, but the images rendered during validation are clearly not the correct poses. Have you ever had a problem like this?
My training and validation results are shown in the figure above.
Thanks for your great work!
I tried to add my own module to BARF, but during the implementation, I found that the model converges normally during training, but the PSNR is very low during validation. Through camera visualization, I can see that the camera poses are all well learned during training, but the images rendered during validation are clearly not the correct poses. Have you ever had a problem like this?
My training and validation results are shown in the figure above.
Thank you very much for your help!