search

OpenDriveLab / ST-P3

[ECCV 2022] ST-P3, an end-to-end vision-based autonomous driving framework via spatial-temporal feature learning.
Apache License 2.0
315 stars 38 forks source link

Dear author ,some question about param #3

Closed emilyemliyM closed 1 year ago

emilyemliyM commented 2 years ago

1)what does "annotation['visibility_token']) == 1" mean? is that means that we will filter the low visibility car?

2)What's more, since the depth images can not be accessed ,If I produce the depth by myself, what's the revolution I should save ? "1600900" or "224480" or any other?

Really thanks.

charleshsc commented 2 years ago

1) It is mentioned in the nuScenes dataset that "The visibility of an instance is the fraction of annotation visible in all 6 images. Binned into 4 bins 0-40%, 40-60%, 60-80% and 80-100%." Yeah, we filter the low visibility car.

2) If you want to produce by yourself, my suggestion is to generate a depth image of the same size as the original image (1600 x 900), so that the original image and the depth image can be kept the same scale during subsequent image processing. You can also generate small depth images(224 x 480) directly if you can ensure that the depth image scale is correct.