-
-
Notifications
You must be signed in to change notification settings - Fork 16.4k
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
two places using multi-scale tricks #21
Comments
@Libaishun yes you are correct! Both methods are suitable for performing multi-scale image augmentation. Exact results depend on model, dataset, etc. |
@glenn-jocher setting multi-scale=True in train.py seems use much more cuda memory than disable it. |
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
speed up pycocotools ops
I see you use multi-scale tricks in two places: directly resizing model inputs in train.py and scale image with random_affine in datasets.py, which one is better to use or shoule we use them simultaneously ?
The text was updated successfully, but these errors were encountered: