Stony coral species recognition system using deep learning
|
| Available in | Authors |
|
Document type: Conference paper
|
| Abstract |
Existing applications for identifying coral species are difficult to find as compared to other artificial intelligence applications, for example, plant recognition applications. People usually recognize coral species by rough observation and identify the information by themselves. It is difficult to recognize the coral species because each species has a wide range of colors, textures, shapes and are very similar to each other, making it rather hard to differentiate unless by trained eyesight. Therefore, a web application system utilizing Convolutional Neural Network (CNN), to identify the name, family name and other related information of a stony coral's species is proposed. 10 Stony coral species were selected for this project. A total of 3831 images were split to train (80%), test (10%) and validate (10%) the model. The model was able to achieve training accuracy of 99%, validation accuracy of 97% and testing accuracy of 91.9% after 50 epochs. The precision, recall and F1 score achieved were 0.88, 0.96, and 0.92 respectively. As a conclusion, the model can classify the stony coral species. This project can be integrated with other platforms in the future. such as mobile applications. |
|
