Pre-trained mode
opened this issue · comments
Can anyone share a pre-trained model I can run test with? I do not have a GPU so I can not try my own mode...
I have the trained model
can you tell me what the box2d in the test command? I mean, for new data, the label of box2d is unknow
You have to have that. For example, using a YOLO. That is the input of this method, I think
I think so! We have to get the 2D detect result. But I find this code in "test" function:
"line[-1] = angle_offset +np.arctan(float(line[11]) / float(line[13]))"
which "line[11]" and "line[12]" and "line[13]"mean the "3D object location x,y,z in camera coordinates (in meters)" in KITTI's object development kit.
however, there are not these information in 2D detect result.
These should be the result of running inference. Test program puts the 2D information and the resulted 3D information in a line and writes them out.
These should be the result of running inference. Test program puts the 2D information and the resulted 3D information in a line and writes them out.
yes，but the 2D information not just the left, top, right, bottom pixel coordinates, the "test" function in this program need the 2D information also include the "3D object location x,y,z in camera coordinates (in meters)"
link：https://pan.baidu.com/s/1XPFqVXYt0SWxrY_bUvhG-w
key：a445
link：https://pan.baidu.com/s/1XPFqVXYt0SWxrY_bUvhG-w key：a445
Thank you for posting a download link. Unfortunately it is not possible to create a baidu account from germany. Therefore I can not download the file. Is it possible to download the checkpoint from any other platform?