search

ApeironY / Modular-Trajectory-Prediction-Toolkit

Codes for paper "Three Steps to Multimodal Trajectory Prediction: Modality Clustering, Classification and Synthesis", "Human Trajectory Prediction with Momentary Observation" and "Stimulus Verification is a Universal and Effective Sampler in Multi-modal Human Trajectory Prediction"
MIT License
31 stars 2 forks source link

Question about test_dataset and val_dataset #2

Open YonghaoDong opened 2 years ago

YonghaoDong commented 2 years ago

Hi,

Thanks for your paper and code for "Three Steps to Multimodal Trajectory Prediction: Modality Clustering, Classification and Synthesis". It's an amazing job.

I want ask a question about val_dataset and test_dataset in your code. It seems you use the same dateset for val_dataset and test_dataset. For example, for eth dataset you use eth_8_12.npy as the validation and test dataset. But you can't use the same dataset for test and validation.

Looking forward to your reply.

GothicAi commented 2 years ago

Thanks! We use this file as default for convenience. Users can get ADE/FDE according to the Evaluation part and corresponding predicted trajectories according to the Testing part. It is free to change the val and test file.