Command for running via docker

[daspk04]

Hello,
I did try to run via docker but it shows error as ''cannot open file" or the "file doesn't exit". I'm not sure if it is the right way to forward input files in the folder in host to docker container. If you could help it would be great.

[remicres]

In you host, you must set the correct rights to the folders and files used inside the container.
What you can do: Start your docker container in interactive mode. Once in your docker container, note the group id (GID) of the folder/file you want to access. Then, create a new group for this GID in your docker container. After that, just add the user (otbuser) in the group owner. Commit your changes (docker commit) then you rights should be Ok.

[daspk04]

Thank you for helping. As you pointed out the problem was with my host folder/files rights as well as I figured out the command for accessing the folder for input was also wrong. So first I gave the folder read, write and execute access.
"sudo chmod -R a+wrx /path/to the/folder". In my case it was
sudo chmod -R a+wrx/home/das/Das_Task

After that I executed all the commands and all worked. Below are the list commands that worked in my case.

Change Directory
pwd: ~/Das_Task/Deep_Learning_Examples
Step 1: Polygon Statistics
sudo docker run -u otbuser -v $(pwd):/home/otbuser mdl4eo/otbtf1.6 otbcli_PolygonClassStatistics -vec /home/otbuser/Data/Pookwe/Shapefiles/Ratoon_Mar1_2018_reprojected.shp -field OBJECTID -in /home/otbuser/Data/Pookwe/Images/SENTINEL2A_20180323-033825-637_L2A_T47QRU_stacked.tif -out /home/otbuser/vec_stats_objid.xml

Step 2: SampleSelection
sudo docker run -u otbuser -v $(pwd):/home/otbuser mdl4eo/otbtf1.6 otbcli_SampleSelection -vec /home/otbuser/Data/Pookwe/Shapefiles/Ratoon_Mar1_2018_reprojected.shp -field OBJECTID -in /home/otbuser/Data/Pookwe/Images/SENTINEL2A_20180323-033825-637_L2A_T47QRU_stacked.tif -instats /home/otbuser/vec_stats_objid.xml -out /home/otbuser/points.shp

Step 3: PatchExtraction
sudo docker run -u otbuser -v $(pwd):/home/otbuser mdl4eo/otbtf1.6 otbcli_PatchesExtraction -source1.il /home/otbuser/Data/Pookwe/Images/SENTINEL2A_20180323-033825-637_L2A_T47QRU_stacked.tif -source1.patchsizex 16 -source1.patchsizey 16 -vec /home/otbuser/points.shp -field OBJECTID -source1.out /home/otbuser/samp_patches_16X16.tif -outlabels /home/otbuser/samp_patches_labels.tif

Change Dircetory
pwd: ~/Das_Task
Step 4: Create Model for Training
sudo docker run -u otbuser -v $(pwd):/home/otbuser mdl4eo/otbtf1.6 python /home/otbuser/OTB_DeepLearn/otbtf-master/python/create_savedmodel_simple_cnn.py --nclasses 2 --outdir /home/otbuser/Deep_Learning_Examples/SavedModels

Step 5: Train Model
sudo docker run -u otbuser -v $(pwd):/home/otbuser mdl4eo/otbtf1.6 otbcli_TensorflowModelTrain -model.dir /home/otbuser/Deep_Learning_Examples/SavedModels -training.targetnodes optimizer -training.source1.il /home/otbuser/Deep_Learning_Examples/samp_patches_16X16.tif -training.source1.patchsizex 16 -training.source1.patchsizey 16 -training.source1.placeholder x -training.source2.il /home/otbuser/Deep_Learning_Examples/samp_patches_labels.tif -training.source2.patchsizex 1 -training.source2.patchsizey 1 -training.source2.placeholder y -model.saveto /home/otbuser/Deep_Learning_Examples/SavedModels/variables/variables

Step 6: Model Serve
sudo docker run -u otbuser -v $(pwd):/home/otbuser mdl4eo/otbtf1.6 otbcli_TensorflowModelServe -source1.il /home/otbuser/Deep_Learning_Examples/Data/Pookwe/Images/SENTINEL2A_20180323-033825-637_L2A_T47QRU_stacked.tif -source1.placeholder x -source1.rfieldx 16 -source1.rfieldy 16 -model.dir /home/otbuser/Deep_Learning_Examples/SavedModels -output.names prediction -out /home/otbuser/Deep_Learning_Examples/outmap.tif uint8

[remicres]

I'm glad this helped! I had a few struggles myself with file permissions inside/outside the docker container.
Just in case: I have updated the README.md with the new docker images (OTBTF r1.7). There is new cuda-enabled docker image, suited for NVIDIA GPUs. A post also some little tricks & tips related to docker containers in this post

[daspk04]

Thank you for the information. The blog was really helpful. At present for some reason my graphic card isn't working. I have GTX 1050Ti may be because it is not compatible with my motherboard or may be faulty although the GPU fan keeps on running. So I wonder if I install the OTBTF r1.7:GPU without graphics card it should also work fine right.? I will not pass the parameter "runtime=nvidia" during docker run. Or should I use OTBTF r1.7:CPU until I have a NVIDIA GPU..?

[remicres]

I think that you can run it without the nvidia runtime, it should behave as the cpu version.
Regarding the GTX1050Ti issue, maybe you need to start from a clean platform with the latest drivers, cuda version, etc.

[daspk04]

Thank you. I did install the OTBTF r1.7:gpu. But I had to create an ''otbuser" inside the container. I did follow the steps from OTBTF r1.6 mentioned as below.

otbtf/tools/dockerfiles/r1.6/bionic.tf-r1.14.otb-develop/Dockerfile

Line 193 in 0aa20d2

# Add user otbuser # Replace 1001 with your user / group id

I wonder why it was removed from the OTBTF r1.7:gpu docker build file. Also in OTBTF r1.6 the working directory for the container was set during build i.e home/otbuser

otbtf/tools/dockerfiles/r1.6/bionic.tf-r1.14.otb-develop/Dockerfile

Line 211 in 0aa20d2

WORKDIR /home/otbuser

but in OTBTF r1.7 it wasn't set. So I have to pass working directory path for each docker run command for example
sudo docker run -u otbuser -w="/home/otbuser/" -v $(pwd):/home/otbuser mdl4eo/otbtf1.7:gpu otbcli_PolygonClassStatistics -vec /home/otbuser/Data/Pookwe/Shapefiles/Ratoon_Jan_Feb_Mar1_reprojected.shp -field P_M -in /home/otbuser/Data/Pookwe/Images/SENTINEL2A_20180323-033825-637_L2A_T47QRU_stacked.tif -out /home/otbuser/Output/1.Polygon_stats/vec_stats_objid.xml

But I can run all the commands. Thank you again.