Meno:Miroslav
Priezvisko:Psota
Názov:Object Detector Trained on Synthetic Data
Vedúci:RNDr. Michal Malý, PhD.
Rok:2020
Kµúčové slová:Machine learning, Object detection, Neural networks, Synthetic data
Abstrakt:Object detection has made a big advancement in the last few years. Object detectors are so advanced these days, that bottleneck of their performance has shifted to the availability of high quality annotated data. Since annotating a large amount of data is sometimes prohibitively expensive, we investigate the scenario, when we have a CAD model of objects of interest available. We use Detectron2 framework and synthetic data generator to train an object detector that is not the best performing, but good enough for bin-picking scenario.

Súbory diplomovej práce:
Autor nedal súhlas so zverejnením svojej diplomovej práce.

Súbory prezentácie na obhajobe:

slides.pdf

Upravi»