perception/object/camera/waitnet/docs/waitnet_mainsection.md

Go to the documentation of this file.

# Copyright (c) 2019-2020, NVIDIA CORPORATION.  All rights reserved.

@page waitnet_mainsection WaitNet

@note SW Release Applicability: This module is available in **NVIDIA DRIVE Software** releases.

## About This Module

This module provides the APIs to initialize, query, and release the NVIDIA
proprietary wait condition object detection deep neural network: **WaitNet**.<br>
WaitNet provides intersection detection, traffic light detection, and traffic sign detection.

The data structures include:
- `dwWaitNetParams`: defines the WaitNet model variant with the specific precision and processor optimization for loading.

DriveWorks comes with two WaitNet models:
- Multiple intersections waypoints, signs lights model: ::DW_WAITNET_MODEL_MULTITASK_WAITNET_V0.
- Multiple intersections signs lights model: ::DW_WAITNET_MODEL_BBOX_INTERSECTION_TL_RS_V1.

The modules are trained with images from 60&deg; and 120&deg; field of view camera.

## Inputs

WaitNet consumes FP16 planar frames with a resolution of 960x604 (multiple intersections signs lights model) or 960x576 (single intersection model) pixels from AR0231 cameras (revision >= 4).

@note Resizing the input frame is internally handled by @ref waitcondition_usecase3 as part of the normal workflow.

## Outputs

WaitNet detects a range of objects related to wait condition. The following three labels (classes) are currently supported:
- Intersections (by multiple intersections signs lights model and single intersection model).
- Traffic signs (by multiple intersections signs lights model).
- Traffic lights (by multiple intersections signs lights).

![WaitNet Outputs](sample_waitnet_perception.png)
Colors indicate the following:
- Orange - Traffic sign.
- White - Traffic light.
- Yellow - Intersection.

### Additional Information

The WaitNet is trained on both day time and night time data.

@warning Currently, the following limitations could affect its performance:
- It is optimized for daytime, clear-weather data. As a result, it
  performs better in daytime with clear weather.
- It is trained primarily on data collected in the United States.
  As a result, it may have reduced accuracy in other locales,<br>
  particularly for road sign shapes that do not exist in the U.S.

## Relevant Tutorials

- @ref waitnet_usecase1.

## APIs

- @ref waitnet_group.