Thank you for your job that combines mono camera 3D object detection with multi-view 3D object detection, I try to implement this method in my own custom datasets.
But when I skim your code, I notice that the configure file of mv-fcos3d++ is for mv-dfm indeed, and the model structure of mv-fcos3d is also a child class of dfm. So I wonder where is the mv-fcos3d++? And do I need to pretrain a dfm following the dfm training process?
Now my dataset can be utilized to train the pdg network.
Thank you for your job that combines mono camera 3D object detection with multi-view 3D object detection, I try to implement this method in my own custom datasets.
But when I skim your code, I notice that the configure file of mv-fcos3d++ is for mv-dfm indeed, and the model structure of mv-fcos3d is also a child class of dfm. So I wonder where is the mv-fcos3d++? And do I need to pretrain a dfm following the dfm training process?
Now my dataset can be utilized to train the pdg network.