Does anyone know if TensorflowLite/Micro has a method to reduce camera resolution. I can get 160 x 120, 2 byte RGB from an Onivision $5 ov7670 camera connected to the Nano 33 BLE, just wondering if anyone knows if Tflite has methods to reduce that to any other resolutions. Such as 32x24, 80x60, black and white, 1 byte color or grayscale. Most languages Python and Javascript have fairly easy ways to resize images, just wondering if any one knows how to do it with either tflite or mbed.
I don’t believe TFLite can resize images, but I’d like to know this as well. I worked through this example: https://github.com/EdjeElectronics/TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi/blob/master/TFLite_detection_webcam.py, and the author uses OpenCV to resize each frame rather than TFLite.
Is there a reason that you can’t use a pooling layer to do the job, at least for integer scalings?
Pooling could definitely work. I don’t know if it’s any more or less computationally efficient than other resizing methods, though.