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

More complex Auto annotation handling? #501

Closed
benhoff opened this issue Jun 17, 2019 · 5 comments
Closed

More complex Auto annotation handling? #501

benhoff opened this issue Jun 17, 2019 · 5 comments

Comments

@benhoff
Copy link
Contributor

@benhoff benhoff commented Jun 17, 2019

Some of the openvino models require a significant amount of post processing. Specifically the text detection model outputs a [1x2x192x320] matrix and a [1x16x192x320] matrix that need to be combined together to create the bounding boxes.

See here for an example of the post processing calculations (a & b are the matrixes talked about).

Is there a good way to do this with the current autoannotation app? I think with the limited keywords the answer is not currently, but wanted to make sure I wasn't missing anything.

@nmanovic

This comment has been minimized.

Copy link
Collaborator

@nmanovic nmanovic commented Jun 18, 2019

@benhoff , probably you are right. We restricted environment for running a python script to avoid security issues. It is just a surprise for me that an output layer so complex. What is your suggestion?

@benhoff

This comment has been minimized.

Copy link
Contributor Author

@benhoff benhoff commented Jun 18, 2019

Is there a reason the security model is like that? One option would be to make only administrators capable of uploading new models, or requiring administrator privileges to execute more privileged code.

There's more structural issues with the code, since the first output is automatically selected in the code base currently. See this line of code where the self._output_blob_name is used to select/return only a single result.

This is a great project, btw! Thanks for all your hard work!

@nmanovic

This comment has been minimized.

Copy link
Collaborator

@nmanovic nmanovic commented Jun 18, 2019

@benhoff , need to think about administrator privileges for such code and don't restrict the python environment. I hope it is a solution.

There's more structural issues with the code, since the first output is automatically selected in the code base currently. See this line of code where the self._output_blob_name is used to select/return only a single result.

The feature is relatively new. Let's improve it together. If you send us a PR I will be happy to accept it. Otherwise definitely we will fix your findings in the future.

@nmanovic nmanovic added this to the Backlog milestone Jun 18, 2019
@nmanovic

This comment has been minimized.

Copy link
Collaborator

@nmanovic nmanovic commented Jun 18, 2019

@benhoff , just clarification. Regular users should be able to upload a model but the model should be run in the restricted environment. For users with admin privileges we can use the python environment without restrictions.

@benhoff

This comment has been minimized.

Copy link
Contributor Author

@benhoff benhoff commented Jun 21, 2019

See: #514

@benhoff benhoff closed this Jun 21, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Linked pull requests

Successfully merging a pull request may close this issue.

None yet
2 participants
You can’t perform that action at this time.