Advancements in Synthetic Video Generation for Autonomous Driving
Figure
2-10 : SEAN normalization . [ 3 ]
UnrealCV : Connecting computer vision to unreal engine proposed by Qiu et al . [ 9 ]. It is an opensource plugin that helps us deal with Game Engine outputs ( Unreal Engine 4 ) 17 . We will see the different possibilities of generating inputs ( Segmented , depth ) for user perspective .
UnrealCV is one of the approaches along with CARLA . It is an open Source simulator powered by Unreal Engine . The generation of test inputs to our proposed system is done using various approaches which includes Unreal Engine integration with ROS Bridge , CarSim 18 , and MATLAB 19 .
Liao et al . [ 7 ] and Cheng et al . [ 8 ] discussed methods for generating inputs that will be useful to our proposed model . We will use these approaches to get the required data , which are advanced methods for generating corresponding outputs . Preprocessing while training any deep neural network is an important aspect .
In this study , we use similar configurations of inputs ( image and semantic segmented ) as proposed in video-to-video synthesis . It employs ideas from image-to-image translation methods for understanding different scenario generation and higher-quality image generation . SEAN seems to be a prime solution in our proposed system ( to use its style feature in multiple ways ) rather than existing SPADE with respect to quality , diversity , and styling aspects .
17 https :// www . unrealengine . com / en-US
18 https :// www . carsim . com / products / carsim / index . php
19 https :// www . mathworks . com / help / matlab / release-notes-R2020a . html 116
March 2024