how are the bounding boxes in the labels read?

[rasoulid]

We are trying to run your code on another dataset with diffrenet dimension images, however, we get wrong bounding boxes. we are assuming you scale down and/or change coordinates of the bounding boxes in the labels, would you please explain how did you use bounding boxes coordinates? Also, when you use the kitti datset the bounding boxes coordinates seems to be decimal, wuold you please explain how you read them? Thank you so much.

[akosiorek]

Hi,

would you please explain how did you use bounding boxes coordinates?

We use KITTI images scaled to 1/2 of the original size and the bounding boxes are scaled accordingly. The format we use is [y, x, height, width].

when you use the kitti datset the bounding boxes coordinates seems to be decimal, wuold you please explain how you read them?

That is correct; why do you think this a problem? The bounding box from the first frame is used to initialise the spatial attention mechanism, which does work with real-valued coordinates.

I hope this helps, otherwise I will need more information about your setup.