Closed YuyaKobayashi closed 5 years ago
(NOT URGENT, because this problem is written in the SDK wiki.)
In current implementation, deeper network reuses input buffer so that input data is corrupted.
Instead of that, it is better to reuse buffer for middle layers.
We run a network for the same input in DEMO application to show FPS.
However, customer may run a network for the same input at only once, probably. Therefore, this is not a problem.
I will close this issue.
(NOT URGENT, because this problem is written in the SDK wiki.)
In current implementation, deeper network reuses input buffer so that input data is corrupted.
Instead of that, it is better to reuse buffer for middle layers.