Question about train dataset and test dataset split

metadriverse / trafficgen

[ICRA 2023] The official code for paper "TrafficGen: Learning to Generate Diverse and Realistic Traffic Scenarios"

Apache License 2.0

153 stars 24 forks source link

Question about train dataset and test dataset split #42

Closed nemonswx closed 6 months ago

nemonswx commented 10 months ago

In the paper, you mentioned '(70000 scenarios after filtering) are further split into a training dataset (50,000 cases) and a non-overlapping test dataset (1,000 cases).' However, I did not see the how you did it in 'init_dataset.py' and 'act_dataset.py'. So how exactly did you split these scenarios?

Alan-LanFeng commented 6 months ago

Hi, we manually split the data into different folders in advance.