View on GitHub

ScanNet

ScanNet Data Server

The ScanNet server is broken down into the following components:

  1. Upload process used by the iPad to upload the scan and trigger scan processing. To ensure that scans can be automatically processed, the scans should be placed in a directory with lots of space and accessible to the scanning processor. Tested on Ubuntu 14.04 Linux machine.
  2. Scan processing scripts. Require a Windows machine with high-end GPU (tested on Windows 10 with GTX Titan X with 12GB of VRAM).
  3. Indexing scripts. Go through scan folders and collate information about the scans.

Upload scripts

The upload script receives scan files (.h264, .depth, .imu, .txt, .camera) from the iPad and stores them in a staging area for scan processing. The files are first placed in the tmp directory before being moved into the staging directory. Uses flask with gunicorn with 10 worker threads on port 8000.

Scan processing scripts

The scan processing requires a Windows machine with high-end GPU (tested on Titan X 12GB). It also relies on these third party tools:

TODO: Also have a brief list of our binaries that are needed and pointers to their repository locations

Indexing scripts

Indexing scripts are used to collate information about the scans and index them.

Statistics computation scripts

Configuration files