jrebut
commented
2 years ago Hi,
Our short term goal is to release the dataset together with the source code to be able for some other to reproduce the paper and play with the data. Indeed, the current version does not include object dimensions. It s not that hard to get the width from the Laser scanner and even the box heading angle, then the height from the camera. For the length of the box, this is harder on a frame by frame basis. We will consider your request and might release an updated label file generated automatically
Julien
Hi authors
Thank you very much for providing such an interesting dataset!
I just wonder if it is possible to include the bird eye's view 2D bounding box ground truth label in the dataset as well? According to your released paper, https://arxiv.org/pdf/2112.10646.pdf, we could see such 2D bounding boxes labels are provided in bird eye's view(i.e. (c) and (d) in figure 1), but we can not find them in the csv file attached in the data. If it is not possible for you to include the bird eye's view 2D bounding boxes label in the dataset recently, could you please provide us some hints on how such bird eye's view 2D bounding boxes labels are drawn in your paper?
Thank you very much again in advance