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.
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]
--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
--batches batches | number of generated batches of size 32
Classifier training
--trn-data-size trn-data-size | number of images for training
--model-name model-name | filename of SVM model