• State Resolved
  • Locked Locked
  • Replies 1 reply
  • Subscribers 97 subscribers
  • Views 273 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TDA2PXEVM: InData parameter in TIDL inference configuration file if numFrames is more than one

Ahmed Anwar
Intellectual 260 points
Part Number: TDA2PXEVM


Hi,

I need to run inference on multiple tensors using quantStatsTool. My input is raw data, so I set rawImage = 1 in the inference configuration file. I've found that numFrames can be set to the number of input tensors to be processed from the input file. If I set numFrames to be greater than 1, how the input data should be provided ?

Based on this thread, I convert a numpy array to a binary file by normalizing between -1 and 1, quantizing the values to signed or unsigned int8 and flattening the tensor. So, In case of only 1 tensor, the input binary file has a flattened quantized tensor. How should be the input file if the number of tensors are more than 1 ? Should I concatenate the tensor values after each other and still have a flattened tensor, or should I add tensors in a second dimension so the binary input file has a 2D tensor ?

Thanks,

Ahmed Anwar