simaaisigmoid

It applies the sigmoid activation function to the output of inference layers, typically in scenarios involving binary classification, confidence scaling, or heatmap normalization. This plugin typically comes after the inference plugin and before post-processing or visualization.

Sigmoid activation, change all data in buffer with the rule:

value = 1 / (1 + std::exp(-value));

To build the plugin use:

source /opt/poky/4.0.10/environment-setup-cortexa65-poky-linux
mkdir build && cd build && cmake .. && make

Installation

Yocto SDK installed.

For manual installation:

scp libgstsimaaisigmoid.so sima@<IP address of DaVinci board>:/data/my_pipeline/libs

Properties

Property	Description
`threshold`	Optional post-sigmoid threshold (e.g., 0.5 to classify as positive)
`input-type`	Specifies input tensor layout or format
`scale` / `bias`	Optional linear adjustments before applying sigmoid

Usage

Example of usage:

GST_DEBUG=2 gst-launch-1.0 --gst-plugin-path=/data/simaai/applications/EfficientDet/lib simaaisrc location=/data/ev_kernel.out delay=1000 ! simaaisigmoid config="/data/config_sigmoid.json"" ! fakesink

Configuration

Configuration example:

{
  "version" : 0.1,
  "node_name" : "sigmoid",
  "simaai__params" : {
    "params" : 3, <============ NOT USED
    "cpu" : 0, <=============== NOT USED
    "next_cpu" : 0, <========== to define the memory type to allocate
    "out_sz" : 8, <====== Output buffer size. Calculated by
    "no_of_outbuf" : 2, <====== Num of output buffers created by the plugin
    "debug" : 0, <============= NOT USED
    "dump_data" : 0 <========== Save output buffer into file
  }
}