Closed devinaconley closed 5 years ago
The memory-leak
branch doesn't exist. Is there a fork that has the reproducible code?
@AnthonyOliveri the branch should be there now
Can this problem be reproduced in the simulator, or does it only occur on a device?
If only on a device, can you give information about the device on which the problem was recreated?
The app requires camera so it does need device.
We got the error on iphone x, 7 plus, 6s
On Sun, Mar 3, 2019 at 12:52 PM Mike Kistler notifications@github.com wrote:
Can this problem be reproduced in the simulator, or does it only occur on a device?
If only on a device, can you give information about the device on which the problem was recreated?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/watson-developer-cloud/swift-sdk/issues/943#issuecomment-469052062, or mute the thread https://github.com/notifications/unsubscribe-auth/ABnY26AY1Y0CUgCxvAlca68A1yXXn1IFks5vTBnigaJpZM4abU41 .
-- सन्जीब िघमिरे (Sanjeev Ghimire)
@mkistler let's discuss as this may still be an open issue.
This issue has been automatically marked as stale because it has had no recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
Description
This problem has only been observed when the device is under very heavy stress, in this case, including high frame rate of classification against WVR Core ML models, along with ARKit usage.
The crash observed is thrown at this line in the SDK when referencing the
model
variable: https://github.com/watson-developer-cloud/swift-sdk/blob/06181082d79d39de6ba6222f8bda8541a920cf95/Source/VisualRecognitionV3/VisualRecognition%2BCoreML.swift#L199The associated error message is:
Reproducing
This crash can be recreated by running the
ar-resume-with-visual-recognition
project on the following branch: https://github.com/IBM/ar-resume-with-visual-recognition/tree/memory-leakVersions:
Solution/workaround:
Current workaround is to throttle frame rate of classification to reduce stress on SDK. Ideally we could provide a solution to boost resiliency, even when under heavy stress.
cc: @dokun1 @sanjeevghimire