Skip to content

Conversation

dmenig
Copy link

@dmenig dmenig commented May 18, 2018

Hi. I spent two days working on this repo in order to render batch evaluation possible. I've created a set of parallel functions (test_images next to test_image, proposal_layer_batch next to proposal_layer etc...).

It isn't made for training, but it can make the utilization of this repo much easier in deployment.

I spent so much time on it I thought I'd suggest the pull to you.

Thanks for your work anyway.

@ruotianluo
Copy link
Owner

Sounds useful. I will take a look. Thank you

@dmenig
Copy link
Author

dmenig commented May 18, 2018

Thanks. i'd be happy to discuss my choices if you have thoughts. For example I had to keep some layers (fc7 for instance) on a sequential evaluation fashion since its inputs can theoretically be of different sizes throughout a single batch.

@ruotianluo
Copy link
Owner

Would you like to extend the training code to support batch too so that we don't need extra *_batch?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants