EMA Project

A automatic metadata extractan project.

• About • Scraper • Notebooks • Processed dataset • Credits

About

EMA project is a contribution towards the development of automated metadata annotation tools. At the current stage a refined dataset is ready to use and has been tested in a multi-label classification model.

Scraper

The scraper directory contains the script needed to download, treat and create the EMA database by thesaurus and labels. To execute, please follow the readme inside the source folder.

Notebooks

The notebooks directory contains the resnet50 execution script and the exploratory part before the model database achievement.

Processed dataset

If you are interested in the dataset generate by default from the scraper script you can use the one submited to the iPres 2022 Conference. Access here.

The dataset have the following structure:

dataset
├── bed
├── bottle
├── candlestick
├── chair
├── chandelier
├── chest
├── chest of drawers
├── cup
├── cup of tea
├── curtain
├── cutlery
├── dish
├── fork
├── glass
├── jug
├── kerosene lamp
├── lamp
├── lantern
├── luminaire
├── mirror
├── oratory
├── pan
├── platter
├── saucer
├── sconce
├── sideboard
├── spoon
├── table
├── table knife
├── vase
└── wardrobe

Credits

Thanks to Amanda Oliveira, from IBRAM for the valuable information she provided regarding her experience in Brazilian museums. We also thank the Tainacan community for their support.