Monitoring the Condition of the Coral Reefs along the Gulf of Eilat: How to Classify and Map Coral Reefs Using Deep Learning

ELI PANOV
Jan 12
2 min read

Updated: Feb 20

Developing an application of the computerized deep learning methodology to recognize corals in a shallow reef in the Gulf of Eilat, Red Sea.

The Challenge: Collecting Underwater Photos for Data Analysis

This project aims to apply deep neural network analysis, based on thousands of underwater images, to automatically recognize some common species among the 100 species reported to be found in the Eilat coral reefs. This is a challenging task, since even in the same colony, corals exhibit significant within-species morphological variability, in terms of age, depth, current, light, geographic location, and inter-specific competition. Since deep learning procedures are based on photographic images, the task is further challenged by image quality, distance from the object, angle of view, and light conditions.

Our mission was to create a model that could:

To produce a large dataset of coral images that were classified into species in the present automated deep learning classification scheme and capture high-quality images of coral reefs.
Train a convolutional neural network (CNN) on labeled images to classify different coral species.
Use techniques like transfer learning with pre-trained models (Resnet) to improve accuracy.

Our Solution: AI-Powered Accurate Coral Identification and Coverage Mapping

We provide an excellent tool for real-time monitoring of the detrimental effects of global climate change and anthropogenic impacts on the coral reefs of the Gulf of Eilat and elsewhere, and that would help assess the success of various bioremediation efforts.

Coral Reef Classification: Key Steps

Taking sufficient high-quality underwater images.
Detecting the chosen coral and cropping its image.
Downscaling the cropped images to 200 × 200 pixels.

Underwater scene featuring diverse corals in shades of yellow and gray. Yellow boxes highlight various coral types . — Figure 1. Image processing in the data set.

From Data to Impact: Uploading Photos into the Dataset

The data collected from videos was transformed into photos and uploaded to the Dataset:

Underwater videos of the coral reef species from the Gulf of Eilat were filmed to produce a large dataset of images.
Image blocks of 200 × 200 pixel-sized image frames were manually cut to comprise the chosen training dataset of some coral species.

Technological Implementation

Methods used in this project include:

Natural sampling of coral reefs by photographing them during the daytime.
Line transects to estimate coral cover percentages at four test sites in the Gulf of Eilat.
Deep convolutional neural networks (CNNs) for accurate classification of coral species using a supervised deep learning (DL) method.

Data Annotation

More than 5,500 images of some coral species were annotated

Computational Techniques Used

Deep learning methods for species classification and image analysis.
Rapid development of accurate coral species classifications.
Precise identification of coral types and coverage.

The Role of this Project

This project aims to:

Identify coral species and monitor the effects of climate change on the coral reefs.
Enable real-time monitoring of changes in reef structure and biodiversity.
Establish a benchmark for future coral research in the region.
Refine and develop DL methods applicable to reefs elsewhere.