wei-mao-2019 / LearnTrajDep

code for learning trajectory dependencies for human motion prediction
MIT License
244 stars 45 forks source link

Problems of the evluation result in the paper LTD for 1000ms prediction #21

Open LaLaLailalai opened 2 years ago

LaLaLailalai commented 2 years ago

Hi,

I try to run the code of LTD (Learning Trajectory Dependencies for Human Motion Prediction Wei) and get the same result as you released in your paper and github repository.

There is a problem with the result when I run "main_3d_eval.py". I want to get the prediction result for 1000ms, so I set the output_n to 25, then I can't get the same result as your paper has written. Also, the short-term (400ms) result contained in 1000ms prediction csv file changed. (If I set output_n to 10, I'll get a similar result. )

Do you know what's the reason for this and how to fix it?