Question about Input T(t - 1)

[HanzhouLiu]

Dear authors,

I really appreciate your great work and contributions to occupancy prediction! In your published paper, Figure 2 shows that there are two inputs Tt and a previous frame T(t-1); however, in codes, I did not see where the temporal fusion was implemented. Also, how did you handle the very first frame, because it does not have any previous frame? Thank you!

[NotACracker]

Dear authors,

I really appreciate your great work and contributions to occupancy prediction! In your published paper, Figure 2 shows that there are two inputs Tt and a previous frame T(t-1); however, in codes, I did not see where the temporal fusion was implemented. Also, how did you handle the very first frame, because it does not have any previous frame? Thank you!

temporal fusion is implemented in the LSSViewTransformerBEVStereo, you can read references BEVDet4D and BEVStereo for more details