Celephais

An application for counting students in photos, predicting their flow from data and assigning them to a set of classrooms.

Installation

Just run

pip install --upgrade .

Dependencies

Celephais depends on keras, scikit-learn, cplex, pandas and opencv-python (all packages are automatically installed).

Usage

Celephais takes its argument from command line, you can list them by runnning celephais -h

usage: Celephais [-h] (--image IMAGE | --xml XML | --json JSON)
                 [--show | --ojson OJSON]
                 (--no-train | --predict-xml PREDICT_XML)
                 [--rooms-json ROOMS_JSON] [--print-score]
                 [--save-imgs SAVE_IMGS]

optional arguments:
  -h, --help            show this help message and exit
  --image IMAGE         the image file or folder in which to detect faces
  --xml XML             the xml file or folder which contains required
                        metadata
  --json JSON           the json file which contains dataset for training
                        (--ojson and --show will have no effect)
  --show                show the detected faces in a window, otherwise just
                        prints the number
  --ojson OJSON         if specified, and if --xml is given, metadata are
                        saved with the given filename
  --no-train            exit before training the model
  --predict-xml PREDICT_XML
                        the xml data (folder or file) of lessons to predict
  --rooms-json ROOMS_JSON
                        the json containing the rooms in which classes will be
                        allocated
  --print-score         use a part of the dataset to calculate the score of
                        the net used for the prediction
  --save-imgs SAVE_IMGS
                        the folder in which images with detected faces will be
                        saved as 'detected_ORIGINAL_FILENAME' (cannot be used
                        with --ojson)

Required files

If --xml is specified, celephais will look for xml files in the following form

<data>
    <subject>Machine Learning</subject>
    <hour>10</hour>
    <day>sunday</day>
    <photo>relative/path/to/image</photo>
</data>

while --json needs a list of dictionaries, for example

[{"day": "tuesday", "subject": "Calculus", "hour": 9, "students": 69} ...

--predict-xml expects the same of --xml apart from the <photo> child.

--rooms-json needs a list of dictionaries containing the name of the room and its capacity

[{"cap": 40, "name": "r00"} ...

About

This application takes its name after one of Lovecraft's tales, Celephais