Identification of Fish Suitable for Consumption with Artificial Intelligence Method Using CNN and SVM Algorithms

E-ISSN : 2807-9035 Volume 5, Number 1, May 2025 https://doi.org/10.47709/brilliance.v5i1.5728 Identification of Fish Suitable for Consumption with Artificial Intelligence Method Using CNN and SVM Algorithms Ramadhanty Riqqa Sunduz1*, Ari Purno Wahyu Wibowo2 1,2 1 Widyatama University, Indonesia , *Corresponding Author Article History: Submitted: 26-03-2025 Accepted: 07-04-2025 Published: 11-04-2025 Keywords: Image Processing; CNN; SVM; Fish Classification; Fish Freshness. Brilliance: Research of Artificial Intelligence is licensed under a Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). ABSTRACT In the fishing industry, determining the freshness of fish is essential since it directly affects the quality of consumption and the viability of items on the market. The purpose of this research is to use Convolutional Neural Network (CNN) and Support Vector Machine (SVM) techniques to create an automated system that can determine the freshness of tilapia fish. Identification is carried out by analyzing the visual characteristics of the fish, in particular skin discoloration, which is the main indicator of freshness. The dataset used in this study was obtained from Kaggle, which covers various conditions of tilapia fish with different levels of freshness. Color conversion into RGB, HSV, and LAB formats to get more accurate color information, image normalization to normalize color intensity, and segmentation to highlight pertinent areas are all part of the pre-processing procedure used to increase the model's accuracy. While SVM is responsible for classifying fish into groups that are either acceptable or unfit for ingestion, CNN is utilized to extract features from fish photos. System testing is carried out by comparing model performance based on classification accuracy. The experiment's findings demonstrated that the CNN and SVM combination could accurately classify the freshness of tilapia fish, however performance was heavily reliant on the input image's quality. It is anticipated that this technology will replace less effective manual techniques, lessen human observational bias, and expedite the industrial fish freshness assessment procedure. With this artificial intelligence-based system, the fishing industry can improve the efficiency and accuracy of the fish sorting process, which can ultimately improve the quality of products consumed by the public. INTRODUCTION Fish is a marine commodity that plays a crucial role as a major source of nutrition for the global population (Ratnasari Desi 2019). In Indonesia, marine fish dominate fish consumption at the household level (Sabarudin Saputra, Anton Yudhana, and Rusydi Umar 2022). The nutritional content of fish, especially protein, is very abundant and essential for the human body. In fact, its protein content exceeds that of beef or chicken, making it a highly recommended option. In addition, fish offers more economical and accessible prices for all walks of life (Styorini, Pratiwi, and Widiasari 2022). However, over time, the quality of the fish can deteriorate substantially. These changes can be seen from alterations in skin color, eye condition, gills, and flesh texture. Enzymatic activity, chemical reactions, and bacterial proliferation are the determinants that trigger these changes. As a result, fish becomes unfit for trade or consumption (Honainah, Romadhoni, and Ato’illah 2022). The freshness level of fish also plays an important role in influencing consumer satisfaction and export sustainability. The manual sorting process is considered less efficient because the volume of fish that must be examined is very large. In addition to potentially causing errors due to human error, this conventional method also requires high operational costs and a long time (Arif Agustyawan 2020). Therefore, the identification of fish freshness needs to be done quickly and precisely, especially in large-scale fish handling (TaheriGaravand et al. 2020). This underlies the development of a fish freshness detection system by utilizing various computational-based methods and electronic technology (Fitriyah, Syauqy, and Susilo 2020). The development of science and technology continues to accelerate along with the progress of the times, to support and facilitate human activities. One of the fields of research that is experiencing rapid development is artificial intelligence (AI) (Asrianda, Aidilof, and Pangestu 2021). The implementation of AI in this context is digital image processing, which is used to analyze the color spectrum as an alternative way of determining the degree of freshness of fish, from fresh, less fresh, to rotten conditions (Sholihin 2021). In image processing, there are various algorithms applied, one of which is the Deep Learning method, especially the CNN. The system can recognize a large number of features in the image as input thanks to this approach. The collected data is then classified to distinguish objects from various categories, such as furniture, fauna, vehicles, and This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 44 E-ISSN : 2807-9035 Volume 5, Number 1, May 2025 https://doi.org/10.47709/brilliance.v5i1.5728 others. This process operates by processing information from the pixel configuration in an image, where each pixel has a value that represents its brightness and color level. However, this method requires a longer duration compared to classical machine learning algorithms (Aditya Dwi Putro Wicaksono ; Henri Tantyoko 2023). On the other hand, the Support Vector Machine (SVM) algorithm is frequently employed in traditional machine learning techniques. This algorithm works by classifying the data into two classes using linear functions in a highdimensional feature space. This process involves searching for an optimal hyperplane (separator plane) that can maximize the margin between two data groups by utilizing kernel functions (Alita, Fernando, and Sulistiani 2020). This study aims to apply image processing methods in comprehensively evaluating the freshness of fish. In particular, this research will concentrate on the analysis of fish skin color through the development of Graphical User Interface (GUI)-based software that is expected to be able to present a clear classification of the feasibility of fish consumption. Thus, this research seeks to present an alternative solution that is more efficient and objective than manual methods in identifying the quality of fish freshness. LITERATURE REVIEW Research conducted by (Sholihin 2021) pioneered the development of an automated system for the identification of fish freshness by analyzing gills images, utilizing Convolutional Neural Network (CNN). The study used 150 gill image data classified into three categories: fresh, deteriorated, and rotten. The results showed an impressive level of accuracy, reaching 100% at the training stage and 97 (...truncated)