How to inference the motion control signal?

[Beau-Yang]

Could you please show a simple demo of predicting the motion control signal? I check the repo codes and paper and understand that I need to set the args.use_car_sensor = True and set the value of key car_info of the inputs.

ADAPT/src/tasks/run_caption_VidSwinBert_inference.py

Line 84 in 35fd7cf

inputs = {'is_decode': True,

But I have no idea what I should use as the value of car_info and the model will predict exactly which signals. Could you give some specific instructions? Thanks a lot.

[jxbbb]

Thank you for your interest. The inference codes of motion prediction have not been released, which would be a possible future job for us. And the args args.use_car_sensor is used for an ablation study in our paper (which locates at TABLE III: Single+).