This repository holds support scripts for running the automated-walk-bike-counter on video files produced by ATSAC traffic cameras.
Currently ATSAC has 600 analog cameras streaming video continuously to their control center. There are up to 12 encoders that can produce digital video from these feeds and place those videos on a single shared drive. However, this shared drive is not (currently) an appropriate place to run the walk/bike algorithm, so we want to copy the files to a place that can run the algorithm, run it, and delete the videos.
The pipeline.py script is intended to be run in two places.
First, an internet-connected location with the shared ATSAC drive mounted.
Second, a GPU compute instance that will run the data.
These places do not need to know anything about each other,
and all communication is mediated through putting files in an s3 bucket.
The dependencies in requirements.txt must be installed into your environment.
On the processing side, this should be the same environment as the walk/bike algorithm.
On the upload side, run
python pipeline.py upload "/path/to/files" "s3://path/to/bucket"where path/to/files is a glob path that matches the files on the shared drive.
On the processing side, run
python pipeline.py process "s3://path/to/files"where "s3://path/to/files" is a glob that matches the files in cloud storage.