Reconstruction of the nadir area of 360 photos using homography to hide the car where the 360 camera is attached to like in Google Street View.
The direction of the sun is automatically calculated using the time and GPS information to choose the side facing the sun to avoid shadows cast by the car.
Vibe coded initially using GitHub Copilot, later with Google Antigravity.
git clone https://github.com/seen-one/NadirCoverUp.git
cd NadirCoverUp
python -m venv env
.\env\Scripts\activate (Windows)
source env/bin/activate (Linux/Mac)
pip install -r requirements.txt
exit
On Windows, exiftool.exe should be installed in the same directory as the script.
Your 360 camera should record at a high enough framerate (15-30 fps) for sufficient tracking.
Configure your inputfolder and maskImagePath in NadirCoverUp.py.
inputfolder should contain an image sequence with output Output_0001.jpg, Output_0002.jpg, etc.
The .gpx file should be placed in the same folder as inputfolder.
If desired, you may edit the other parameters to suit your needs.
.\env\Scripts\activate (Windows)
source env/bin/activate (Linux/Mac)
python NadirCoverUp.py
The script will create a new folder at OutputStep6 with the processed images, ready for upload to Mapillary or Panoramax. If wanting to upload to Google Street View, you can use ffmpeg to encode a video from OutputStep5 image sequence then upload the video along with the .gpx file.