Giters

lmb-freiburg / Multimodal-Future-Prediction

The official repository for the CVPR 2019 paper "Overcoming Limitations of Mixture Density Networks: A Sampling and Fitting Framework for Multimodal Future Prediction"

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

EMD on CPI dataset

mprzewie opened this issue · comments

commented

Hi!

Could you specify the way (or even better, the code you've used) to evaluate the EMD metric on the CPI Dataset? The only EMD-related stuff in your code seems to be related with the SEMD metric calculated for the SDD results.

Thanks in advance! :)

commented

Hi @mprzewie

Please have a look at the new uploaded file https://github.com/lmb-freiburg/Multimodal-Future-Prediction/blob/master/emd_cpi.py

It is a simple script which loads the ground-truth samples from CPI and the prediction files containing the parameters of mixture distribution and compute the EMD distance.

Note that we used 1000 samples from the ground-truth distribution (i.e, you need to generate the CPI test data by running the python CPI-test.py output_folder num_test_scenes 3 1000).
For the prediction, we expect three files .float3 containing the parameters of the predicted mixture model (means, sigmas, and weights) with shapes ([4,2], [4,2], [4]) assuming the predicted mixture model has four components.

Best,

commented

Thank you so much!