This patch adds an 8-bit grayscale mode that halves the image buffer needed by the MCU and presents the image data in the format the TensorFlowLite Micro person detection example expects.
Grayscale mode is selected in a similar way to the existing RGB565 or YUV422 color formats:
Camera.begin(QVGA, GRAYSCALE, 1)
Note - the grayscale option is in fact requesting YUV422 from the camera, but is disposing the U and V bytes and only storing Y to the data array in a Camera.readFrame(dataArray) call - so only 1 byte/pixel is required. HT @sandeepmistry
This patch adds an 8-bit grayscale mode that halves the image buffer needed by the MCU and presents the image data in the format the TensorFlowLite Micro person detection example expects.
Grayscale mode is selected in a similar way to the existing RGB565 or YUV422 color formats:
Camera.begin(QVGA, GRAYSCALE, 1)Note - the grayscale option is in fact requesting YUV422 from the camera, but is disposing the U and V bytes and only storing Y to the data array in a
Camera.readFrame(dataArray)call - so only 1 byte/pixel is required. HT @sandeepmistry