search

ChuhuaW / SGNet.pytorch

Pytorch Implementation for Stepwise Goal-Driven Networks for Trajectory Prediction (RA-L/ICRA2022)
117 stars 16 forks source link

Failed to reproduce results and Bugs in related Trajectron++ code #19

Open MeiliMa opened 2 years ago

MeiliMa commented 2 years ago

I found that your implementation uses an old version of Trajectron++ code with bugs during data reprocessing (see StanfordASL/Trajectron-plus-plus#53). I tried to reproduce the reported results using the last code of Trajectron++ with that bug fixed. The following are the results I achieved:

    ADE/FDE (reproduced)    ADE/FDE (reported)
eth:   0.47/0.77 ------------> 0.35/0.65
hotel: 0.20/0.38 ------------> 0.12/0.24
univ:  0.33/0.62 ------------> 0.20/0.42
zara1: 0.18/0.32 ------------> 0.12/0.24
zara2: 0.15/0.28 ------------> 0.10/0.21

I used your implementation with the default configs to reproduce the results. Could you confirm if there is anything wrong during my reproducing the results?

ChuhuaW commented 2 years ago

Hi @MeiliMa,

Thank you for bringing it to our attention! We were unaware of the Trajectron++ bug when we conducted our experiment. And thank you for conducting new experiments! I'll look into it and get back to you later.