Closed SwagJ closed 1 year ago
For the annotation of moving objects, we employed the following steps:
I'm sure that there are better ways. In particular, the paper by Xieyuanli Chen from our group, see also the repository for some code, proposes an automated combination of ground removal, HDBSCAN clustering, and then tracking to associate instances over time. This could be a starting point to then correct the labels.
Hope that helps.
Hi @jbehley,
Thank you for much for your detailed illustration. That completely solves my problem.
Best,
Hi @jbehley,
Thank you for open-source this amazing tool. I have one question regarding to the method to annotate moving object. If multiple scans are aggregated across the whole frame, then moving objects(such as cars, pedestrain) will be quite noisy right? How would you handle this? Thank you in advance. I am looking forward to your reply.
Best,