search

princeton-vl / DPVO

Deep Patch Visual Odometry/SLAM
MIT License
583 stars 66 forks source link

NAN appears when training #54

Open JesseZZZZZ opened 3 months ago

JesseZZZZZ commented 3 months ago

Hi @lahavlipson , have you tried training DPVO on other datasets beside TartanAir? I tried to train it on TUM (using _build_dataset from raw RGB and depth maps), but all values became NAN, do you know the issue? (t1[0] and t2[0] are all NANs). Thank you so much! ![Uploading image.png…]()

JesseZZZZZ commented 3 months ago

Hi, dear authors, it turns out that the nomalization step will cause NANs in poses image After I delete the two lines of codes like I did in the picture, the issue disappeared. Do you think that simply deleting the codes is available? Or are there any other ways to solve this issue? THANKS!