COCO (Creating Common Object in Context) Dataset for Chemistry Apparatus
Silvia Rostianingsih, Alexander Setiawan, Christopher Imantaka Halim
Abstract
In order to create machine learning, we need to build a model. The model is created from a process called training. The goal of training is to develop an accurate model that answers some questions and in order to train a model, we need to collect a dataset. The quality and quantity of the data gathered will determine how good the predictive model can be. Helping the model to understand datasets like humans do is one of the important processes of machine learning. Datasets need to be constructed and transformed correctly. In this research, we compare the difference between creating a COCO dataset manually and creating a synthetic COCO dataset. Creating datasets for chemistry apparatus is not as difficult as creating a human object. The apparatus has a specific shape and form, thus the dataset had to have a limited number. As a result, we create both a dataset both manually and synthetically. The synthetic dataset helps to gain more datasets by combining some objects with different backgrounds.