refiner
Attributes
Classes
A Python based gstreamer plugin template. Enables the user to: |
Module Contents
- refiner.plugin_name = 'refiner'
- class refiner.MyPlugin
A Python based gstreamer plugin template. Enables the user to: - Accept incoming buffers from dynamic pads - Define any custom plugin runtime logic User has to only override the run() function
- mla
- fmap1_shape = [1, 80, 192, 256]
- fmap2_shape = [1, 80, 192, 256]
- netlist0_shape = [1, 80, 192, 128]
- netlist1_shape = [1, 40, 96, 128]
- netlist2_shape = [1, 20, 48, 128]
- inp_list0_shape = [1, 80, 192, 384]
- inp_list1_shape = [1, 40, 96, 384]
- inp_list2_shape = [1, 20, 48, 384]
- coords0_shape = [1, 80, 192, 2]
- coords1_shape = [1, 80, 192, 2]
- flowup_shape = [1, 80, 192, 1]
- upmask_shape = [1, 80, 192, 144]
- coords0_shape_aligned = [1, 80, 192, 8]
- coords1_shape_aligned = [1, 80, 192, 8]
- flowup_shape_aligned = [1, 80, 192, 8]
- fmap1
- fmap2
- net_list_0
- net_list_1
- net_list_2
- inp_list_0
- inp_list_1
- inp_list_2
- coords0
- coords1
- unpack_shapes
- run(input_buffers: List[python_plugin_template.SimaaiPythonBuffer], output_buffer: bytes) None
Define your plugin logic HERE Inputs: input_buffers List[SimaaiPythonBuffer]: List of input buffers Object of class SimaaiPythonBuffer has three fields: 1. metadata MetaStruct Refer to the structure above 2. data bytes - raw bytes of the incoming buffer 3. size int - size of incoming buffer in bytes
- unpack_buffer(single_buffer)
- preprocess(l_image, r_image)
- run_model(fmap1, fmap2, net_list_0, net_list_1, net_list_2, inp_list_0, inp_list_1, inp_list_2, coords0, coords1, iter)
- preprocess_data_refiner(tensor)