I’m working on an application that detects objects with detection+tracker model (mediapipe) and sets anchors on detected objects (ARCore and Sceneform).
Current implementation:
- Acquire
Image
from ARCoreFrame
inScene.OnUpdateListener.onUpdate
method - Copy image buffer to byte array and convert to RGB Bitmap using RenderScript
- Pass it to Mediapipe graph through
FrameProcessor.onNewFrame(bitmap, timestamp)
Is there any way to avoid copying and converting image to bitmap and pass ARCore GPU texture to Mediapipe graph directly?