Hi, you said that: "input data is a 4-channel BGRG raw unnormalized Bayer data as in this tutorial ("RAW_combined" will be used instead of "RAW_norm"):
https://github.com/MediaTek-NeuroPilot/mai21-learned-smartphone-isp/blob/master/load_dataset.py"
Cause the 'RAW_norm' is a np.float32 tensor, so I want to know the dtype of your input 'RAW_combined' is np.float32\ np.int32 or np.uint16. If it is uint16, how can we transform an uint16 pb model to tflite model
Posted by: rhwang @ March 18, 2021, 9:46 a.m.