search

aimotive / aimotive_dataset

aiMotive public dataset
https://openreview.net/forum?id=LW3bRLlY-SA
Other
44 stars 2 forks source link

Syncronization problem #5

Closed dav695 closed 9 months ago

dav695 commented 10 months ago

Hello, Congratulations on this fantastic work! I am trying to use your dataset to test some fusion architectures. But I project the pointcloud lidar in the image, and sometimes I see a mismatch. Like this two examples in the motorbike (20210920-203858-00.22.38-00.22.53@Sogun):

sync2 pr1

Or this example in the highway (20220203-144759-00.40.00-00.40.15@Cali):

problem_front_img

I think that the matrix transformations are correct. Do you know why this things happen? Thanks you so much!

TamasMatuszka commented 9 months ago

Hi @dav695,

We updated our paper with the limitations that explain the mismatches: "The sensor setup results in a synchronization limitation caused by the fact that sensors differ in the temporal recording method. Since the rotating LiDAR and rolling shutter cameras have different measurement methods, some discrepancies exist in temporally discretized annotations. This phenomenon is mostly visible when the relative speed difference between the ego and exo car is large." (https://arxiv.org/pdf/2211.09445.pdf)

However, the last case cannot be explained by this fact. We will investigate the problem. Thanks for letting us know!

Since our baseline algorithms operate in BEV space, these synchronization issues did not cause significant problems for us. I hope the mismatches won't prevent you from utilizing the dataset.

dav695 commented 9 months ago

Hello @TamasMatuszka , thanks you so much for the anmswer. I close the issue, and again congratulations!!