search

yuantianyuan01 / StreamMapNet

GNU General Public License v3.0
186 stars 16 forks source link

temporal fusion #30

Closed myabest closed 1 month ago

myabest commented 1 month ago

I have a question, at line 173 of StreamMapNet.py:

Backbone

_bev_feats = self.backbone(img, img_metas=img_metas, points=points) It seems that the prev_bev parameter (previous bev featues) is not given here. Does it mean that timing fusion is not done later? At the same time, I noticed that the is_first_frame parameter is always True. Why is this?

yuantianyuan01 commented 1 month ago

Hi,

The temporal fusion is done at line 177, where we use a memory bank to store and update the bev features. The is_first_frame should not always be true. Please make sure you are using the correct config file, for example https://github.com/yuantianyuan01/StreamMapNet/blob/main/plugin/configs/nusc_newsplit_480_100x50_24e.py.

myabest commented 1 month ago

thank you~