Working with ROIs

Regions of interest (ROIs) are polygonal areas which can be tested for the presence of objects or other events such as:

  • If an object enters or leaves a region.

  • The time an object stays inside a region.

  • The number of objects in a region.

  • The amount of movement or activity inside a region.

  • Etc.

Combined with the information from an object detector, an object tracker, or motion estimation, ROIs can provide useful insights of what is happening in a scene.

As an example Fig. 4 shows how regions are applied to a scene with a fixed camera to associate the presence of human operators or pixel activity in each region to different tasks being performed.

_images/mirtar-rois.png

Fig. 4 Regions in the deck of a ship for classification of on board tasks in video footage.

API reference

JSON format for definition of polygonal regions

{
   "regions": [
      {
         "name": "tangon_babor_sup",
         "polygon": [[385, 315], [676, 334], [754, 625], [668, 810], [165, 757]],
         "color": [155,155,0]
      },

      {
         "name": "tangon_babor_inf",
         "polygon": [[11, 532], [777, 606], [666, 1073], [5, 1074]],
         "color": [155,155,0]
      },

      {
         "name": "tangon_estribor_sup",
         "polygon": [[1057, 336], [908, 682], [954, 773], [1522, 754], [1565, 595], [1210, 305]],
         "color": [155,0,155]
      },

      {
         "name": "tangon_estribor_inf",
         "polygon": [[973, 1072], [949, 625], [1749, 538], [1904, 1068]],
         "color": [155,0,155]
      },

      {
         "name": "puente",
         "polygon": [[1088, 369], [592, 382], [570, 580], [1131, 584]],
         "color": [0,255,0]
      }
   ]
}

This module contains components for loading and processing of polygonal regions of interest (ROIs).

class videoanalytics.pipeline.sinks.roi.ROIPresenceCounter(name, context, filename, context_name='DETECTIONS')

Component for testing the presence of detected objects in ROIs.

This component READS the following entries in the global context:

Variable name

Description

START_FRAME

Initial frame index.

This component WRITES the following entries in the global context:

Variable name

Description

q_{name}

Number of objects inside region {name}.
Parameters

  • name (str) – the component unique name.

  • context (dict) – The global context.

  • filename (str) – name of JSON file containing region definitions.

  • context_name (str) – name of the variable in the context containing the detections.

(*) The entry contains a list of dictionaries containing:

  • polygon: numpy array containing polygon definition

  • color: color to represent the polygon in the video.

process()

This method is called for each active component in the pipeline.

setup()

This method is called after all components from the pipeline are instanced.

shutdown()

This method is called after the process finished.

class videoanalytics.pipeline.sinks.roi.ROIView(name, context, filename, alpha=0.2)

Component for visualization of ROIs on frame.

This component READS the following entries in the global context:

Variable name

Description

START_FRAME

Initial frame index.

This component WRITES the following entries in the global context:

Variable name

Description

regions(*)

Region definitions.
Parameters

  • name (str) – the component unique name.

  • context (dict) – The global context.

  • filename (str) – name of JSON file containing region definitions.

  • alpha (float) – transparency factor (0=invisible,1=opaque).

(*) The entry contains a list of dicitonaries with following elements:

  • polygon: numpy array containing polygon definition

  • color: color to represent the polygon in the video.

process()

This method is called for each active component in the pipeline.

setup()

This method is called after all components from the pipeline are instanced.

shutdown()

This method is called after the process finished.