Setup Instuctions

You must have an updated version of Blender installed to render new track layouts.

Open a terminal and enter the following commands:

git clone https://github.com/RufusVijayaratnam/SLAM.git

cd SLAM

pip install -r requirements.txt

Note: If on Windows, then PyTorch cannot be install normally.

• Go to the PyTorch website and choose the relevant options to install PyTorch.
• CUDA is not necessary but will speed up YOLOv5 training.

Drawing a new track

In the SLAM directory:

cd 'Draw and Annotate'

To draw a new track called "example"

python DrawTrack.py example

A window will open, you must sequentially click to add points representing the track layout. The green dot represents the starting point.

Press "esc" to finish drawing.

Rendering

Open Blender Environment called FSD Environment.blend located in the Blender directory.

Click Window > Toggle System Console to open the console.

Navigate to the scripting tab and select the ConePlacement.py.001 script and set the following variable on line 15:

track = "example.txt"

Run the script.

The track will be imported into the Blender environment.

Render a video

To render the sequential images for a full video render, select the RenderVideo.py script, set the following variable on line 9:

track = "example.txt"

Run the script.

Blender will become unresponsive, the console will show the progress of the track render.

Create a video

• You must have FFmpeg installed.

In the SLAM directory:

cd Blender/Resources/Renders/video/images

To create an .mp4 file for each stereo camera:

ffmpeg -framerate 30 -i example-Right_Cam-Render-%04d.png -pix_fmt yuv420p ../../Videos/example-right.mp4

Replace example with the name of your track.

Repeat the command but replace Right_Cam with Left_Cam