About the test dataset:only 20 points?

[coco-99-coco]

Thanks for your work first!However,I found that each track in the test dataset has only 20 historical track points.In this way,after I have trained the model on the traindataset and adjusted it with the valdataset ，I can't test it on the testdataset.How can I use the testdataset?

[johnwlambert]

In Section 5.1 of our paper, we define the motion forecasting problem as:

The forecasting task is framed as: given the past input coordinates of a vehicle trajectory V_i = (X_i, Y_i)
where X_i = (x_i^t, y_i^t) for time steps t = {1, . . . , T_obs}, predict the future coordinates
Y_i = (x_i^t, y_i^t) for time steps {t = T_{obs+1}, . . . , T_{pred}}. For a car, 5 s is sufficient to capture
the salient part of a trajectory, e.g. crossing an intersection. In this paper, we define the motion
forecasting task as observing 20 past frames (2 s) and then predicting 30 frames (3 s) into the future. 

The last 30 timesteps are held out as test data on a sequestered test set (with evaluation accessible via the EvalAI leaderboard). Please see https://github.com/argoai/argoverse-api/issues/155.