Basically I'm working on a react-native app that uses the react-native-vision-camera library for the camera component. The library provides an API for writing native frame processors on the native side of each platform (ios and android). I'm trying to use tensorflow lite to make inferences over the frames the camera reads. On ios I got it working pretty easily but on android I can't find any efficient way to pass the image that I get from the frameprocessor callback to the tensorflow model.
The frameprocessor callback gets the frame as input and I can extract an android.Image from it or an hardwareBuffer. I'm thinking hardwarebuffer would be the most efficient way to work with the frame.
Would it be possible to pass the hardwarebuffer to a tflite model on android?