SEASaw provides users the ability to textually-search for video based on the visual contents of the video.
- clone of this repo
- (for database access) a valid credential file from us
- (for database access) a valid database password from us
- python 3.5, pip
- linux
- current working directory at root of repo
- chmod -x cloud_sql_proxy.amd64
- pip3 install -r requirements.txt --user
- nltk's tokenizer, stopwords, punctuations import nltk nltk.download('stopwords', '/home/vagrant/nltk_data') nltk.download('punkt', '/home/vagrant/nltk_data')
python3 -m seasaw.start
This will launch the application in its simplest mode, allowing for you to use the frontend to search through existing data. Will require the below args to function fully.
- --gca_credentials_path will need to be a path to the credential file
- --database_password for the database password
python3 -m seasaw.scheduler
This command requires the previous one to be running, this will run the indexer.
- --gca_credentials_path will need to be a path to the credential file
- --database_password for the database password
- all previous prerequisites for running without scraper (except OS requirement), plus
- (for scraper to run) a valid imgur password from us
- hashicorp vagrant (to make life easier) - https://www.vagrantup.com/
- virtualbox (for vagrant)
- vagrant up
- vagrant ssh
- cd /vagrant
- python3 -m seasaw.start
Run the start with the following args:
- --gca_credentials_path will need to be a path to the credential file
- --database_password for the database password
- --imgur_password with the imgur password
- -s to enable the scraper
- -l known as "local mode", ie, set all internal urls to work on the vagrant box.
Commands (cont...)
- python3 -m seasaw.scheduler
- --gca_credentials_path will need to be a path to the credential file
- --database_password for the database password
If this is your first time running the project, vagrant up may take some time, as it will be downloading dependencies
When finished, be sure to tear everything down with "vagrant halt", your battery will thank you.
Once spinning, you will be able to access the frontend through port 25285 ("/")
The API for the datasource can be found at port 25280 or 25281 ("/doc")