AUTO City Deep Learning Dataset Plugin Overview
This is a plugin developed by Vertechs based on Unreal Engine 4 that is useful in efficiently improving object detection and recognition of autonomous driving AI.
AUTO City Deep Learning Dataset Plugin Features
Parameters, output format, etc. can be set according to the intended use and verification environment.
Simultaneous output of camera sensor, segmentation, and depth image is possible from the same environment.
Playback recording simplifies scenario creation operations.
.png)
Contents
Playback Recording
It is possible to create a scenario while operating the vehicle you want to move on UE4. Compared to conventional scenario creation, scenarios can be generated without requiring complicated operations.
Simultaneous generation of camera, segmentation, and Depth images
It is possible to output the camera, segmentation, depth image, and each image at the same time from one environment. It is possible to completely match the reproducibility between each image.
Segmentation
The segmentation image identified by AI at the pixel level can be easily output with a few actions without complicated processing.
Depth image
Depth images that express images in depth can be output on an 8-bit basis without any special processing.
AUTOCITY PLUGIN usage
Generate segmentation, depth with playback recording feature, and output the camera images from Autocity. These segmentation colors are not based on KITTI, these can be modified as per your requirements of testing. The above graph shows the connection between camera image output and the evaluation target.