[Motion Forecasting Dataset] Insufficient ego-vehicle(object_category = 3) samples in some scenarios from training dataset

argoverse / av2-api

Argoverse 2: Next generation datasets for self-driving perception and forecasting.

https://argoverse.github.io/user-guide/

MIT License

319 stars 76 forks source link

[Motion Forecasting Dataset] Insufficient ego-vehicle(object_category = 3) samples in some scenarios from training dataset #59

Closed Mitchell-Lee-93 closed 2 years ago

Mitchell-Lee-93 commented 2 years ago

Thank you for the great work.

I'm dealing with motion forecasting data and I found out that, from my understanding, some scenarios from training data have insufficient ego-vehicle(object_category = 3) samples. For example, 'scenario_00baa620-0342-4b0c-a7c2-77a5f583d6b1' only have 100 samples for agent instead of 110.

I'm wondering if my understanding of the motion forecasting dataset is wrong or if I should ignore the problematic scenarios

Mitchell-Lee-93 commented 2 years ago

Sorry. Didn't check the formal issue.