Skip to main content

4.1.2 Importing the Model Inference Library

The hobot_dnn model inference library is pre-installed on the Ubuntu system of the development board. Users can import the module and check the version information.

sunrise@ubuntu:~$ sudo python3
Python 3.8.10 (default, Mar 15 2022, 12:22:08) 
Type "help", "copyright", "credits" or "license" for more information.
>>> from hobot_dnn import pyeasy_dnn as dnn
>>> dir(dnn)
['Model', 'TensorProperties', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__spec__', 'load', 'pyDNNTensor']

The main classes and interfaces used in the hobot_dnn inference library are as follows:

  • Model: AI algorithm model class, used for loading algorithm models and performing inference calculations. For more information, please refer to the Model documentation.
  • pyDNNTensor: AI algorithm input and output data tensor class. For more information, please refer to the pyDNNTensor documentation.
  • TensorProperties: Class for the properties of the input tensor of the model. For more information, please refer to the TensorProperties documentation.
  • load: Load algorithm models. For more information, please refer to the API interface documentation.