search

jchengai / forecast-mae

[ICCV'2023] Forecast-MAE: Self-supervised Pre-training for Motion Forecasting with Masked Autoencoders
https://arxiv.org/pdf/2308.09882.pdf
145 stars 15 forks source link

ensemble code #10

Open LL-C8 opened 6 months ago

LL-C8 commented 6 months ago

Hi! Can you provide the experimental code for Ensemble?

jchengai commented 6 months ago

Hi, @LL-C8. We use k-means for endpoint clustering and linearly re-scale the probability. I think the implementation is quite straightforward.

LL-C8 commented 4 months ago

Hi, @LL-C8. We use k-means for endpoint clustering and linearly re-scale the probability. I think the implementation is quite straightforward. Could you please provide me with the combinations of learning_rate and attention radius that you selected when choosing the submodels? Thank you!