What data are we supposed to use for validation and submission?
Posted by: Avhirup @ June 27, 2018, 11:28 a.m.The data links are available in https://competitions.codalab.org/competitions/18423#participate. You can find submission instructions there.
devX.zip is validation set without labels. In this phase you submit your predictions on validation set and can check how accurate your model is. Then you can tune your parameters. In a few days we will release validation labels.
I am having this error while trying to run process.py
/bin/sh: 1: th: not found
Traceback (most recent call last):
File "/home/interns/anaconda3/envs/tf/lib/python3.5/shutil.py", line 544, in move
os.rename(src, real_dst)
FileNotFoundError: [Errno 2] No such file or directory: 'joints.dat' -> '../../Y/joints.dat'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "process.py", line 19, in <module>
shutil.move("joints.dat",args.SubmissionFolder+"/joints.dat")
File "/home/interns/anaconda3/envs/tf/lib/python3.5/shutil.py", line 558, in move
copy_function(src, real_dst)
File "/home/interns/anaconda3/envs/tf/lib/python3.5/shutil.py", line 257, in copy2
copyfile(src, dst, follow_symlinks=follow_symlinks)
File "/home/interns/anaconda3/envs/tf/lib/python3.5/shutil.py", line 120, in copyfile
with open(src, 'rb') as fsrc:
FileNotFoundError: [Errno 2] No such file or directory: 'joints.dat'
I ran the following command. where X is the input,devY is my results and Y is an empty folder
python Track1/processing/process.py X devY Y
It seems like it is not finding torch..
Does the "th" command do anything? if not, are you running this script through the docker container? if you are trying to run this from outside the container you will have to install torch.
Also probably not related to your problem, but you may want to run the process.py script from the processing folder to avoid other errors.
No, i am not running it in from a container. Ya that's the issue, I would have to install torch to execute the commands.
Posted by: Avhirup @ July 4, 2018, 6:04 p.m.Hello , i am having trouble getting to make the process script to work, after install torch, it required me to install hdf5 and cudnn . Unfortunately, i dont have sudo access to machines i use to train models, so can't install it on that machine, and trying to get it to work on my laptop(with no gpu) does not work . Isn't there some other way out? maybe a python script or some other way to submit?
Posted by: Avhirup @ July 5, 2018, 8:12 a.m.Hi,
You can use docker container we provided. You can use Dockerfile we provided in starting kit to install within container. Within docker you have root permission.
Best,
Posted by: mmadadi @ July 5, 2018, 9:26 a.m.Hello!
Sadly, we do need torch as we use it for the human pose estimation, without it we cannot predict the WNJD. As mmadadi already said the best option would be to install Docker on the computer you used to train, our container already comes with torch + all the dependencies needed to run the script.
As a last resort you can download this makeshift script https://drive.google.com/file/d/1YlJBVXYzoHDLUsWztH36k0czeB3s-fHN/view?usp=sharing, which will allow you to submit without human pose estimation, your WNJD score will be meaningless, but you will be able to upload the submissions and check the other metrics.