Open OctoberKat opened 2 years ago
Hi, we have preliminarily tried using positional encoding, which did not help in our case. My understanding is that the scene flow representation for autonomous driving scenes might not necessarily need positional encoding, since we want to make the local region "as rigid as possible". Cheers.
Hi, thanks for your great work! I wonder if you have tried using positional encoding or sine activation schemes in the coordinate-MLP network? It is a very popular method to make the coordinate-MLP fitting high frequencies signals better, but in your source code it seems that the coordinate-MLP is a relu-activated MLP without positional encoding.