SakuraRiven

https://github.com/open-mmlab/OpenPCDet/blob/master/pcdet/datasets/nuscenes/nuscenes_dataset.py#L143 Could we add ```'gt_boxes' in data_dict:``` in above line to avoid key_error? Because I want to just inference some middle sweep frames without annos.

I think the ```nums_epochs_to_eval``` means a positive number to count evaluated epochs. However it starts with 0 which may cause confusion. Basically, when it equals 1, the last epoch is...

关于性能

1

你好，请问densenet有跟直接使用原作者的类vgg做比较吗，效果上会不会更好一点，有没有资料可以参考下？

the difference of iou3d between target_assign and nms

1

Hi, we noticed that the iou3d used in [target_assign](https://github.com/tianweiy/CenterPoint/blob/db36c497a71014961c1ec17042a7524a79d4e792/det3d/ops/iou3d_nms/iou3d_nms_utils.py#L62) contains z-axis which is reasonable. However, the iou3d in [nms](https://github.com/tianweiy/CenterPoint/blob/db36c497a71014961c1ec17042a7524a79d4e792/det3d/core/bbox/box_torch_ops.py#L270) is only the bev perspective. Are there something we missed? Thanks...

iou3d_nms does not contain z-axis

3

Hi, we noticed that the ```iou3d_nms``` only emply calculate iou on the bev perspective without height. Why do we drop the iou on z-axis, and could such nms improve the...

Hi, we want to create __atomicMax__ function with __half inputs by following [pytorch_scatter/atomics.cuh at master](https://github.com/rusty1s/pytorch_scatter/blob/master/csrc/cuda/atomics.cuh). The detailed codes are inline __device__ void operator()(at::Half *address, at::Half val) { \ unsigned int...

The histogram comparison and DUDE dataset

5

Thanks for sharing such great work! I have some questions and would appreciate your kind help: (1) The original metrics[1] of different methods on BIndingDB dataset are fuzzy histogram. How...