Splat Creation

• download pictures from google bucket using CLI
• convert to jpg if needed
• Download folder using "gsutil -m cp -r gs://bucket-name "{path to local existing folder}""
• Upload your Data to the cluster using scp (scp -r /path/to/your/images/ username@servername:/path/to/save/your/images/)
• connect to the cluster using ssh
• find a unoccupied pool using "pool-smi"
• Activate Nerfstudio:
  • source ~/.bashrc
  • conda activate nerfstudio
• start the preprocessing using COLMAP ( ns-process-data {images/video} --data /path/to/save/your/images/ --output-dir /new/path/to/processed/images/)
• check if the pool is still unoccupied (pool-smi)
• Start the training (ns-train splatfacto --data /new/path/to/processed/images/ --output-dir ./outputs)
• Export the splat (ns-export gaussian-splat --load-config outputs/...[experiment_path].../config.yml --output-dir ./exports/splat)
• Dowload the splat.ply file (scp -r username@servername:./exports/splat/splat.ply /.../hyperrealistic_indoor_streetview/)
• Convert the .ply file to a .splat file using node ./convert_ply_to_splat.js (or just upload it to [Supersplat](https://playcanvas.com/supersplat/editor and export as .splat)