PRBonn / lidar-bonnetal

Semantic and Instance Segmentation of LiDAR point clouds for autonomous driving
http://semantic-kitti.org
MIT License
960 stars 206 forks source link

[Errno 2] No such file or directory: '/c\u200b\u200bonfig/arch/darknet21.yaml' #14

Closed windtries closed 5 years ago

windtries commented 5 years ago

sorry,when i train it,it occurs such a mistake

tano297 commented 5 years ago

Can you share your setup? Python version, Linux version, command you run, etc.

windtries commented 5 years ago

Can you share your setup? Python version, Linux version, command you run, etc.

I use python3.7.1,Linux version 4.15.0-65 and the command is: ./train.py -d home/Harddisk_6T/gouxiaofeng/lidar-bonnetal/train/tasks/semantic/dataset -ac /c​​onfig/arch/CHOICE.yaml -l home/Harddisk_6T/gouxiaofeng/lidar-bonnetal/train/tasks/semantic/dataset/log Thank you!

tano297 commented 5 years ago

You are trying to look for the config file in the root directory.

Try with a dot to look in the current directory:

-ac ./c​​onfig/arch/CHOICE.yaml

windtries commented 5 years ago

You are trying to look for the config file in the root directory.

Try with a dot to look in the current directory:

-ac ./c​​onfig/arch/CHOICE.yaml

I've tried it,and the same error occurs again.

tano297 commented 5 years ago

Can you please share where you are in the computer, and if those files exist in those paths?

$ pwd
$ ls -al
$ ls home/Harddisk_6T/gouxiaofeng/lidar-bonnetal/train/tasks/semantic/dataset
$ ls ./c​​onfig/arch/CHOICE.yaml
windtries commented 5 years ago

您能否分享您在计算机中的位置以及这些路径中是否存在这些文件?

$ pwd
$ ls -al
$ ls home / Harddisk_6T / gouxiaofeng / lidar-bonnetal / train / tasks / semantic / dataset
$ ls ./c onfig / arch / CHOICE.yaml

yeah ,I find that my path is wrong and I correct it,but maybe the problem is "CHOICE.yaml",it doesn't exist itself.I have tried to change the command with writing the yaml file such as the "darknet21.yaml"which truly exist ,the same error occurs again.

jbehley commented 5 years ago

CHOICE.yaml is a place holder. You have to use a yaml file from config/arch, such as config/arch/darknet21.yaml.

To clarify what is needed: $ ./train.py -d <dataset path> -ac <configuration file> -l <log directory>

<dataset path> = directory of the root directory of SemanticKITTI, i.e., let say you extracted the zip files into /home/Harddisk_6T/gouxiaofeng/semantic-kitti/dataset/ then this folder contains the following folder structure:

/home/Harddisk_6T/gouxiaofeng/semantic-kitti/dataset/
  \-- sequences
        |-- 00
        |     |-- velodyne
        |     |-- labels
        |     |-- calibration.
        |     \-- poses.txt
        |-- 01
       ...

It's not the dataset folder in lidar-bonnetal/train/tasks/semantic.

<configuration file> = this is the configuration file of the architecture. (As also mentioned in the README.MD it might be a file from the config\arch\ directory. (Example would be config/arch/darknet21.yaml) You can modify these files or write your own, but it has to contain the model configuration and the training parameters. An empty file will not do.

<log directory> = is a directory/folder where you want to have the log files and checkpoints stored. (This is also the directory you have to open with tensorboard to get feedback on the progress of training.) This can be an abitrary directory.

You seem to struggle with relative and absolute paths, see https://www.geeksforgeeks.org/absolute-relative-pathnames-unix/ and https://opensource.com/article/19/8/understanding-file-paths-linux As mentioned in the last article Tab is your friend.

Hope that helps.

tano297 commented 5 years ago

Were you able to solve this? I am closing this since it does not look like an issue with the framework. Let us know if you need further help