search

robot-learning-freiburg / PanopticBEV

Bird's-Eye-View Panoptic Segmentation Using Monocular Frontal View Images. http://panoptic-bev.cs.uni-freiburg.de
http://panoptic-bev.cs.uni-freiburg.de
123 stars 22 forks source link

Detection perform very bad after changing input image resolution #11

Open lygbuaa opened 2 years ago

lygbuaa commented 2 years ago

I change the "scale=1.0" to "scale=0.5" in nuscenes.ini, and re-trained the model, expecting similar results, but the object-detection perform very bad in the smaller scale.

result from "scale=1.0", fairly good. image

but result from "scale=0.5", nothing detected ! image

how should I tune the model after changing the scale, any suggestion?