UnIT extended 2019   ItAintMuchButItsHonestWork

You can acces UnIT extended 2019 hackaton assignment here. This repository contains source code created by group ItAintMuchButItsHonestWork. Based on tournament metrics we've reached succes rate of 76.89 % in recognizing ellipses and won first place.

Pipeline

Usage

Application

python3 src/main.py [--mode mode] [--csv-input csv-input] [--csv-output csv-output] [--images-path images-path] [--ground-truths-path ground-truths-path]

Parameters:

--mode mode                              |  application mode (eval, entry)
--csv-input csv-input                    |  filename of input csv file containing training data
--csv-output csv-output                  |  filename of output csv file containing competition data
--images-path images-path                |  image folder path
--ground-truths-path ground-truths-path  |  ground truths folder path

Data generator

python3 src/batch.py [--batches batches]

Parameters:

--batches batches  |  number of generated batches of size 32

Classifier training

python3 src/classification.py [--trn-data-size trn-data-size] [--model-name model-name]

Parameters:

--trn-data-size trn-data-size  |  number of images for training
--model-name model-name        |  filename of SVM model

---

Created by: Petr Buchal, Martin Ivančo, Vladimír Jeřábek