How to replicate the results in your paper

StanfordASL / Trajectron-plus-plus

Code accompanying the ECCV 2020 paper "Trajectron++: Dynamically-Feasible Trajectory Forecasting With Heterogeneous Data" by Tim Salzmann*, Boris Ivanovic*, Punarjay Chakravarty, and Marco Pavone (* denotes equal contribution).

MIT License

664 stars 188 forks source link

How to replicate the results in your paper #78

Closed xiaxx244 closed 1 year ago

xiaxx244 commented 1 year ago

Hi, first of all. This is an interesting paper and I really like the idea of trajectron++. I am currently working on a project related to that and I need to replicate the results in your paper. However, the visualization results and the quantitative evaluation results I obtained is quite different from your paper. I am wondering what is your training parameters and epochs and how you train your model? Can you release your trained model and the train and test split publicly ? Thanks!

BorisIvanovic commented 1 year ago

Hi @xiaxx244 , If you absolutely, at all costs need to replicate the model from the 2020 paper, then take a look at the eccv2020 branch: https://github.com/StanfordASL/Trajectron-plus-plus/tree/eccv2020

Having said that, you can (and should) feel free to use the much more updated repository here: https://github.com/NVlabs/adaptive-prediction (just make sure not to use any of the adaptive features if you only want to compare to T++).