The author's code is really standardized and clear, but I have a little doubt about the output of the model. That is, KITTI officially stipulates that the output value range of alpha and rotation_y should be in [-pi..pi], but the alpha and rotation_y output by the model are obviously not in this range, as shown in the figure below.
Is there any problem in the code part of printing alpha and rotation_y in the model? Does this have an impact on the accuracy of the detection?
The author's code is really standardized and clear, but I have a little doubt about the output of the model. That is, KITTI officially stipulates that the output value range of alpha and rotation_y should be in [-pi..pi], but the alpha and rotation_y output by the model are obviously not in this range, as shown in the figure below. Is there any problem in the code part of printing alpha and rotation_y in the model? Does this have an impact on the accuracy of the detection?