Skip to content

Latest commit

 

History

History
50 lines (38 loc) · 2.48 KB

reference.md

File metadata and controls

50 lines (38 loc) · 2.48 KB
Raw

Purpose

The DS side of Scribble Stadium has a somewhat larger repo compared to other projects. As such, we wanted to make a quick reference guide to help make the transition easier. We hope you find this helpful in your Scribble adventures!

Contents

  • app
  • assets
  • data
  • data_management
  • models
  • notebooks
  • structured_experiments
  • tesstrain

app

The api folder houses the endpoints that comprise the DS API:

  • clustering: This endpoint takes a list of cohort and submission objects and then returns clusters based on cohort in groups of four.
  • db: database functions : This gets a SQLAlchemy database connection. Uses this environment variable if it exists: DATABASE_URL=dialect://user:password@host/dbname. Otherwise uses a SQLite database for initial local development.
  • models: This contains endpoints for text and illustration submissions, along with line graph, histogram, and crop cloud requests.
  • submission: Contains a function that takes a submission object and calls the Google Vision API to text annotate the passed s3 link, then passes those concatenated transcriptions to the SquadScore method. Also contains a function that checks the illustration against the Google Vision SafeSearch API and flags if explicit content is detected.
  • visualization: This contains endpoints for Story Squad’s main visualizations: line graphs, histograms, and the crop cloud.

The app/tests and app/utils folders house the tests and utility functions the API utilizes.

assets

This folder is where some DS architecture and examples are housed in the .png format

data

This folder contains the base image data and Scribble Stadium image data that Tesseract uses to train its OCR model

data_management

The main folder is autopreprocess_testing which houses the preprocessing that was attempted by BloomTech learners.
The other files in the directory are needed for the photo transformer story_photo_transformer.py, which is used in
an experiment in the structured_experiments files

models

This houses the different models tried throughout the project's lifetime. The storysquadalldata4.traineddata
did the best; however, it still needs more work to be used on children's handwriting

notebooks

Research was performed here in the different .ipynb files

structured_experiments

Contains the experiments that Learners performed

tesstrain

Has a modified tesseract Makefile for use in docker