Could you please elaborate on how the provided detections were obtained? Specifically, I am interested in the data used to train the model. There are some conflicting answers in the issues for this repo as
in #36 it is said the model is trained only on the object dataset,
but in #53 it is said that it was trained on the KITTI tracking sequences as well.
Could you please clarify which one it is?
This is an important issue because if the detector is trained on the tracking training dataset, which as far as I understand is the same as the tracking validation dataset, then the object detector was trained on the same sequences as the ab3dmot tracker was evaluated on.
This is a followup on issues #16, #36 #57 and #53 (to which I replied with this same question)
Thank you for work and code, I very much enjoyed using them :)
Vassili
Hi Xinshuo,
Could you please elaborate on how the provided detections were obtained? Specifically, I am interested in the data used to train the model. There are some conflicting answers in the issues for this repo as
Could you please clarify which one it is?
This is an important issue because if the detector is trained on the tracking training dataset, which as far as I understand is the same as the tracking validation dataset, then the object detector was trained on the same sequences as the ab3dmot tracker was evaluated on.
This is a followup on issues #16, #36 #57 and #53 (to which I replied with this same question)
Thank you for work and code, I very much enjoyed using them :) Vassili