Currently imvoxel net is trained on kitti to make only one prediction for the class category.However the dataset has more than just one class such as
pedestrain,cyclist,van,truck,tram and sitting pedestrain.
I tried changing the configuration and the kitti_eval file in mmdet3d/evaluation/functional/kitti_eval.py but I am running into some errors
The changes i made are
Change no of classes in imvoxelnet config to 6(pedestrain,cyclist,car,van,tram and sitting_pedestrain)
Change class_to_name in eval.py at line 701.
Doing these changes results in an error
incorrect dimension of min_overalps index out of error tried to access index 5 of array with axis shape 2.
Can you suggest what changes can I make to the configuration of imvoxelnet so that it can be trained on kitti dataset.
If changes are not possible,can you recommend another model for 3d object detection.
Any other context?
tried working with models like BEV fusion but they are trained on nuscenes which tends to be take up too much storage.
the min only has train and validation files,but i still need to download the test zip which is almost 54GB
What is the feature?
Currently imvoxel net is trained on kitti to make only one prediction for the class category.However the dataset has more than just one class such as pedestrain,cyclist,van,truck,tram and sitting pedestrain. I tried changing the configuration and the kitti_eval file in mmdet3d/evaluation/functional/kitti_eval.py but I am running into some errors The changes i made are
Can you suggest what changes can I make to the configuration of imvoxelnet so that it can be trained on kitti dataset.
If changes are not possible,can you recommend another model for 3d object detection.
Any other context?
tried working with models like BEV fusion but they are trained on nuscenes which tends to be take up too much storage. the min only has train and validation files,but i still need to download the test zip which is almost 54GB