Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Deepdetect crashes often without any warning #45

Closed
shashi-netra opened this issue Jan 9, 2016 · 4 comments
Closed

Deepdetect crashes often without any warning #45

shashi-netra opened this issue Jan 9, 2016 · 4 comments
Assignees
Labels

Comments

@shashi-netra
Copy link

I get an aborted (core dumped) message every so often and Deepdetect stops. I am using it to tag images with the imagenet model

Sometimes the issue is obvious such as a very large image or a dead link, but often it is just a normal image for no discernible reason.

@beniz beniz self-assigned this Jan 9, 2016
@beniz
Copy link
Collaborator

beniz commented Jan 9, 2016

Can you post the typical server log that leads to such a crash, along with the predict POST single or batch call ?
Also, if you are not running the currrent master branch, it would be useful if you could specify which commit you're on.
EDIT: also needed: whether you are using a GPU and what memory it has (or machine RAM is running on CPU)

@beniz beniz added the type:bug label Jan 9, 2016
@shashi-netra
Copy link
Author

Thanks for your reply.
I pulled from the master about 1 month ago for my current install (commit: bda8633)
After looking into the issue some more, it seems the issue is with throwing errors leading to unhandled exceptions and causing the OS to call terminate()

This happens with image formats not supported by OpenCV (such as .gif):

curl -X POST "http://xxx.com:8080/predict" -d "{\"service\":\"imgnet\",\"parameters\":{\"input\":{\"width\":224,\"height\":224},\"output\":{\"best\":5}},\"data\":[\"http://i.ytimg.com/vi/0vxOhd4qlnA/maxresdefault.jpg\"]}"

works fine, but:

curl -X POST "http://xxx.com:8080/predict" -d "{\"service\":\"imgnet\",\"parameters\":{\"input\":{\"width\":224,\"height\":224},\"output\":{\"best\":5}},\"data\":[\"https://upload.wikimedia.org/wikipedia/en/3/39/Specialist_Science_Logo.gif\"]}"
terminate called after throwing an instance of 'dd::InputConnectorBadParamException'
what():  no data for image https://upload.wikimedia.org/wikipedia/en/3/39/Specialist_Science_Logo.gif
curl: (52) Empty reply from server
[1]+  Aborted                 (core dumped) ./deepdetect/build/main/dede -stderrthreshold 0 --host xxx.com > dede.txt

Similarly, broken links cause a crash:

terminate called after throwing an instance of 'dd::InputConnectorBadParamException'
what():  no data for image http://i.ytimg.com/not_a_real_link.jpg
curl: (52) Empty reply from server
[1]+  Aborted                 (core dumped) ./deepdetect/build/main/dede -stderrthreshold 0 --host xxx.com > dede.txt

Thanks in advance for your help.

@beniz
Copy link
Collaborator

beniz commented Jan 10, 2016

Thanks for the thorough report. The crash with .gif can be reproduced so I'll clear this asap.

Note that the .gif are not handled at the moment (nor for training or prediction) since they are made of multiple frames. What do you expect from classifying a .gif, getting a prediction over the first frame or a prediction of each of the frame ?

@beniz
Copy link
Collaborator

beniz commented Jan 10, 2016

Commit above fixes the regression following parallelized image reading. Thanks for reporting this one.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

2 participants