Owen-Liuyuxuan
commented
1 year ago If you are using Ground aware module. It does not generalize really well to different shape. You could try to visualize the reshaped image and detection results on the jupyter notebook.
Perhaps you need to carefully tune the output size.
Hello, i want to inference the Ground-aware Monocular 3D Object Detector on NuScenes dataset. The nuscenes dataset is consisted of RGB images of shape (900,1600,3) and its own calibration matrices. I use your ROS code for inference with the nuscenes data but the cars are not detected. I also changed the rgb_shape in the config file to (900,1600,3) but still it didnt work. Do you know what else should i change in your config file? Thank you in advance!