:py:mod:`uavfpy.odcl.inference`
===============================

.. py:module:: uavfpy.odcl.inference

.. autoapi-nested-parse::

   
   ..
       !! processed by numpydoc !!


Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   uavfpy.odcl.inference.BBox
   uavfpy.odcl.inference.Target
   uavfpy.odcl.inference.TargetInterpreter
   uavfpy.odcl.inference.Tiler




Attributes
~~~~~~~~~~

.. autoapisummary::

   uavfpy.odcl.inference._EDGETPU_SHARED_LIB
   uavfpy.odcl.inference.TENSOR_ORDERS


.. py:data:: _EDGETPU_SHARED_LIB
   

   
















   ..
       !! processed by numpydoc !!

.. py:data:: TENSOR_ORDERS
   

   
















   ..
       !! processed by numpydoc !!

.. py:class:: BBox(xmin, ymin, xmax, ymax)

   Bases: :py:obj:`object`

   
















   ..
       !! processed by numpydoc !!
   .. py:method:: overlap(self, other)

      
















      ..
          !! processed by numpydoc !!


.. py:class:: Target(id, score, bbox)

   Bases: :py:obj:`object`

   
















   ..
       !! processed by numpydoc !!

.. py:class:: TargetInterpreter(model_path, label_path, cpu, thresh, order_key='mobilenet')

   Bases: :py:obj:`object`

   
















   ..
       !! processed by numpydoc !!
   .. py:method:: get_labels(self, label_path)

      
















      ..
          !! processed by numpydoc !!

   .. py:method:: make_interpreter(self, model_path_or_content, device=None, delegate=None)

      
      Make new TPU interpreter instance given a model path


      :Parameters:

          **model_path_or_content** : str
              filepath to model. recommended to use absolute path in ROS scripts

          **device** : str, optional
              None -> use any TPU
              ":<n>" -> use nth TPU
              "usb" -> use USB TPU
              "usb:<n> -> use nth USB TPU
              "pci" -> use PCI TPU
              "pci:<n> -> use nth PCI TPU

          **delegate** : loaded TPU Delegate object, optional
              supercedes "device" flag

      :Returns:

          tflite.Interpreter
              the interpreter













      ..
          !! processed by numpydoc !!

   .. py:method:: load_edgetpu_delegate(self, options=None)

      
      load edgetpu delegate from _EDGETPU_SHARED_LIB with options


      :Parameters:

          **options** : dict, optional
              TPU options, by default None

      :Returns:

          loaded Delegate object
              the TPU













      ..
          !! processed by numpydoc !!

   .. py:method:: input_tensor(self)

      
      get input tensor



      :Returns:

          tensor
              the input tensor













      ..
          !! processed by numpydoc !!

   .. py:method:: set_input_tensor(self, image, resize=False)

      
      set the input tensor from (cv2) image array of size (h, w c)


      :Parameters:

          **image** : np.array
              h, w, c














      ..
          !! processed by numpydoc !!

   .. py:method:: output_tensor(self, i)

      
      Return output tensor regardless of quantization parameters


      :Parameters:

          **i** : int
              which output tensor to grab

      :Returns:

          tensor
              output tensor













      ..
          !! processed by numpydoc !!

   .. py:method:: input_image_size(self)

      
      Get interpreter input size



      :Returns:

          tuple of int
              (height, width, colors)













      ..
          !! processed by numpydoc !!

   .. py:method:: interpret(self, img, resize=True)

      
















      ..
          !! processed by numpydoc !!

   .. py:method:: get_output(self, score_threshold)

      
      Return list of detected objects


      :Parameters:

          **score_threshold** : float
              number from 0-1 indicating thresh percentage

      :Returns:

          list of Target
              list of namedtuples containing target info













      ..
          !! processed by numpydoc !!


.. py:class:: Tiler(size: int, offset: int)

   Bases: :py:obj:`object`

   
















   ..
       !! processed by numpydoc !!
   .. py:method:: get_tiles(self, raw_shape)

      
















      ..
          !! processed by numpydoc !!

   .. py:method:: tile2board(self, tbbox: BBox, wl, hl)

      
















      ..
          !! processed by numpydoc !!

   .. py:method:: merge_overlapping(self, targets: list)

      
      probably can optimize this
















      ..
          !! processed by numpydoc !!

   .. py:method:: parse_localTarget(self, target, wl: int, hl: int)

      

















      ..
          !! processed by numpydoc !!