add database entry to setup script

[mosqueteiro]

pre-requirements:

• [x] git clone project
• [x] EBS volume with dataset mounted to project git folder #26
• [x] docker environment built with pyscopg2 #25

Tasks:

• [x] check database structure is correct
• [x] change segmentation type to BYTEA in initdb.sql
• [x] add BYTEA seg_dims to imgaes table in initdb.sql
• [x] change bbox to BYTEA in initdb.sql
• [x] update BuildDatabase:
  • [x] build_sql don't need create_tables anymore...
  • [x] load_json change "segmentation" to bytestring
  • [x] load_json make "seg_dims" to bytestring
  • [x] load_json change "bbox" to bytestring
• [x] update QueryDatabase:
  • [x] read segmentation from bytestring
  • [x] read seg_dims from bytestring reshape segmentation
  • [x] read bbox from bytestring
• [ ] run get_images on small query table to make sure file pathing setup correctly (no downloading)
• [ ] update data_pipline.py to use argparse in "__main__" to load annotations
• [ ] add bash command to setup script to load appropriate dataset+annotations

[mosqueteiro]

AUDIBLE: going to store sementation as json string to preserve different structures. We won't be using this during current experiments anyways.

• [x] change segmentation type to JSON
• [x] remove seg_dims
• [x] check number of colunms (7)
• BuildDatabase:
  • [x] pop info into self.info
  • [x] convert tables to DataFrames
  • [x] update bbox conversions to bytestring
  • [x] update segmentation conversions to json string
  • [x] write tables to .csv (s) and compare sum of sizes to json size
• QueryDatabase:
  • [x] remove read seg_dims
  • [x] update read segmentation
• [ ] ? move csv(s) to docker/postgres then:
  • [ ] update postgres Dockerfile to copy csv(s) into db

[mosqueteiro]

AUDIBLE: going to store sementation as json string to preserve different structures. We won't be using this during current experiments anyways. ....

possible issue with index of pandas DF, use set_index('column')

[mosqueteiro]

finally have models training again