You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, we find that your model using [-1.1, 0, 0.5, 1.1] as the bev_planes' presetting heights. And your former work HDMapNet using only height 0.
But by our visualization, we find that using raw camera images as input to get BEV results with height -2 can get better BEV results(when height as -2, we can get clean BEV image with strong road semantic information, heights as yours setting the BEV image can't be found road information by eyes).
So we are curious about that there are better heights setting for IPM(for nuscenes data)?
The text was updated successfully, but these errors were encountered:
VectorMapNet_code/configs/vectormapnet.py
Line 82 in 4bcc8b4
Hi, we find that your model using [-1.1, 0, 0.5, 1.1] as the bev_planes' presetting heights. And your former work HDMapNet using only height 0.
But by our visualization, we find that using raw camera images as input to get BEV results with height -2 can get better BEV results(when height as -2, we can get clean BEV image with strong road semantic information, heights as yours setting the BEV image can't be found road information by eyes).
So we are curious about that there are better heights setting for IPM(for nuscenes data)?
The text was updated successfully, but these errors were encountered: