Part Number: TDA4VM
In the multi-batch scenario (eg. numBatches = 10),
1. In the situation of rawData as inputs (the value of inFileFormat is 1 and the input bin file contains the preprocessed data of 10 images), the quantification process in importer can be performed in multiple batches normally.
2. In the situation of multiple images as inputs (the value of inFileFormat is 2), when we dump the output data of each layer of quantification process in importer, It can be found that only the first batch of the input layer is filled with input data, and the data of other batches is 0.
Out question is: Whether tidl model importer support multiple batches modes when the input data is image? If only the first batch has data, whether the quantization precision of the tidl bin model will be affected?
