some issues about process of generate virtual points

[zcj223311]

Firstly, thank you for providing the source code. I want to replicate the process of generating virtual points from you, but I have encountered some difficulties. I have installed the code according to the steps you provided and executed it on the terminal:

python tools/create_data.py nuscenes_data_prep --root_path=NUSCENES_TRAINVAL_DATASET_ROOT --version="v1.0-trainval" --nsweeps=10 --virtual True

This section generates two empty virtual point cloud folders and infos_val_10sweeps_withvelo_filter_True.pkl and infos_train_10sweeps_withvelo_filter_True.pkl file

Execute again: python virtual_gen.py --info_path data/nuScenes/infos_train_10sweeps_withvelo_filter_True.pkl MODEL.WEIGHTS centernet2_checkpoint.pth

He reminded me: No such file or directory: 'home/zcj/CenterPoint/data/nuScenes/samples/LIDAR_TOP_VIRTUAL/n015-2018-07-24-11-22-45+0800__LIDAR_TOP__1532402927647951.pcd.bin.pkl.npy'

I guess I may need to train an image semantic segmentation result first, but I couldn't find any relevant training in your code. Do I need to provide it myself? I also couldn't find a place in the previous code to insert the training results. At the same time, I noticed that I downloaded a pth file while executing the second code mentioned above. What is the purpose of this file.

Also, I am using the mini dataset, and I am not sure if this is the reason

I am a freshman with a limited understanding of this research field and a serious lack of code proficiency. If you could provide more detailed operational procedures, I would greatly appreciate it.

首先，谢谢您提供的开原代码。 我想要复现您生成虚拟点的过程，但是我遇到了一些麻烦。 我按照您给的步骤安装好了代码，并在终端执行了： python tools/create_data.py nuscenes_data_prep --root_path=NUSCENES_TRAINVAL_DATASET_ROOT --version="v1.0-trainval" --nsweeps=10 --virtual True

这一段生成了两个空的虚拟点云文件夹，和infos_val_10sweeps_withvelo_filter_True.pkl与infos_train_10sweeps_withvelo_filter_True.pkl文件。

再执行了： python virtual_gen.py --info_path data/nuScenes/infos_train_10sweeps_withvelo_filter_True.pkl MODEL.WEIGHTS centernet2_checkpoint.pth

这段他提示 No such file or directory: 'home/zcj/CenterPoint/data/nuScenes/samples/LIDAR_TOP_VIRTUAL/n015-2018-07-24-11-22-45+0800__LIDAR_TOP__1532402927647951.pcd.bin.pkl.npy'

我猜测我可能需要先训练一个图片语义分割结果来，但在您的代码中没有找到相关的训练，这是需要我自己提供吗，我也没在前面代码中找到可以插入训练结果的地方，同时我注意到在执行上面提到的第二个代码时下载了一个pth文件，请问这个文件的作用是什么。

另外我用的是mini数据集，不知道会不会是这个原因

我是一个研一新生，对这个研究领域一知半解，代码水平也严重不足，如果作者能够提供更加详细的操作过程，我将万分感谢。

[XxxuLimei]

把‘virtual_gen.py’的269行改为output_path = os.path.join('/', *tokens[:-2], tokens[-2]+"_VIRTUAL", tokens[-1]+'.pkl.npy')即可