pretrained weights about backbone?

[gp1234567]

1. Did you release the pre training weight of backbone？
2. I used two frames of data to superimpose and input on the 128 line lidar. The results of 1.3W data after 20epoch are as follows. Can you give me some comments?

[Abyssaledge]

1. We can not release the pretrained model due to the license of Waymo.
2. Could you present more details about your experiment? Which config do you use? Any modifications?

[gp1234567]

I use sst_v2 backbone + detect_head of anchor free，like centerpoint

[Abyssaledge]

I think the pillar size (0.16m) is too small for CenterHead, because the object center is hard to be completely filled with several convolutions after the SST backbone.
I suggest use pillar of 0.32m and follow the recently updated configs/sst_refactor/sst_waymoD5_1x_3class_centerhead.py.
Let me konw if you make some progresses.

[Abyssaledge]

Could you show me the performance and config of the CenterPoint model? And it would be nice to upload the config file instead of directly posting the config here.

[gp1234567]

@Abyssaledge
Hi，can you give me some comments? thanks.

[Abyssaledge]

Could you upload the complete config of SST model?