search

abduallahmohamed / Social-STGCNN

Code for "Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction" CVPR 2020
MIT License
483 stars 141 forks source link

How to evaluate in different setting? #40

Closed haofanwang closed 3 years ago

haofanwang commented 3 years ago

I have trained a model with obs_len=12 and pred_len=24, how can I test this model under a different setting? For example, obs_len=8 and pred_len=12.

abduallahmohamed commented 3 years ago

Thanks for reaching out. I think for longer predictions than what it was trained for you might need to retrain, but for shorter predictions you can trim [:18] the extra predicted locations before calculating the FDE/ADE metrics while leaving the TrajectoryDataset settings the same. Yet, I still prefer to train a new model for each variant settings as we have an edge in terms of FDE metric.

Abduallah Mohamed abduallahmohamed.com

On Wed, Nov 25, 2020 at 3:16 PM Haofan Wang notifications@github.com wrote:

I have trained a model with obs_len=12 and pred_len=24, how can I test this model under a different setting? For example, obs_len=8 and pred_len=12.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/abduallahmohamed/Social-STGCNN/issues/40, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFIVLGYY27PJOQV7ERLU2N3SRVX3DANCNFSM4UC4436A .