search

agrimgupta92 / sgan

Code for "Social GAN: Socially Acceptable Trajectories with Generative Adversarial Networks", Gupta et al, CVPR 2018
MIT License
813 stars 261 forks source link

Eval results not the same as in the paper? #59

Closed Mirorrn closed 5 years ago

Mirorrn commented 5 years ago

Hi, these are my results from the evaluate_model.py.

Dataset: eth, Pred Len: 12, ADE: 0.70, FDE: 1.31 Dataset: eth, Pred Len: 8, ADE: 0.58, FDE: 1.14 Dataset: hotel, Pred Len: 12, ADE: 0.48, FDE: 1.02 Dataset: hotel, Pred Len: 8, ADE: 0.36, FDE: 0.72 Dataset: univ, Pred Len: 12, ADE: 0.56, FDE: 1.18 Dataset: univ, Pred Len: 8, ADE: 0.33, FDE: 0.70 Dataset: zara1, Pred Len: 12, ADE: 0.34, FDE: 0.69 Dataset: zara1, Pred Len: 8, ADE: 0.21, FDE: 0.42 Dataset: zara2, Pred Len: 12, ADE: 0.31, FDE: 0.64 Dataset: zara2, Pred Len: 8, ADE: 0.21, FDE: 0.42