Title

ECUHO1 (Edith Cowan University Halophila Ovalis 1)

Author Identifiers

MD Moniruzzaman

ORCID: 0000-0003-1130-7078

Collection Type

Dataset

Upload Date

2019

School or Research Centre

School of Science

Contact

Syed Mohammed Shamsul Islam: Shams.webmail@gmail.com

Description

ECUHO1 contains 2699 underwater 2D digital images of Halophila ovalis (a type of seagrass). Images are collected from both natural shallow shorelines and experimental laboratory setup.

Natural shorelines include Dampier reserve, Mandurah, shallow coastlines of Fremantle and Rottness Island, Coral Bay, Exmouth of Ningaloo area from North-Western Australia, Lucky Bay, Milyu, Pelican Point and Rockey Bay of Swan Canning area of WA. The experimental laboratory setup is situated at Edith Cowan College, Joondalup.

The whole dataset is labelled with expert annotation. The image file format is JPG and labelling was saved as XML files. The entire dataset is divided into a test set and a train set. In the annotation files (.csv) for test set and training set contains the following information:

Filename: Name of the Image file.

Width: Width of the whole image (pixel).

Height: Height of the whole image (pixel).

Class: Object Class of the image patch inside bounding box (bbox).

Xmin: Location of the bottom left corner of the bbox.

Ymin: Location of the top left corner of the bbox.

Xmax: Location of the bottom right corner of the bbox.

Ymax: Location of the top right corner of the bbox.

FoR Codes

0801, 0502

DOI

10.25958/5dd7566dfaabb

Research Activity Title

Seagrass Detection Using Deep Learning

Research Activity Description

The Research project was undertaken as a part of masters by research (J16). The project aimed to demonstrate the possibility and effectiveness of deep learning-based object detection techniques for seagrass detection from underwater images. Underwater seagrass (Halophila ovalis) images were collected from both natural shallow shorelines and ECU experimental facility. Collected images were labelled with expert advice. Finally, Faster R-CNN object detection framework with a range of deep neural network architectures has been trained with the dataset.

Start of data collection time period

2017

End of data collection time period

2019

Language

eng

Codes

Filename: Name of the Image file.

Width: Width of the whole image (pixel).

Height: Height of the whole image (pixel).

Class: Object Class of the image patch inside bounding box (bbox).

Xmin: Location of the bottom left corner of the bbox.

Ymin: Location of the top left corner of the bbox.

Xmax: Location of the bottom right corner of the bbox.

Ymax: Location of the top right corner of the bbox.

File Format(s)

JPG, XML

File Size

20 GB

This document is currently not available here.

Share

 
COinS