Identifikasi Pemalsuan Bubuk Cabai Merah Menggunakan Citra Digital Berbasis EfficientNetV2 dengan Visualisasi Explainable AI

Srimaharani, Nabilah Putri (2026) Identifikasi Pemalsuan Bubuk Cabai Merah Menggunakan Citra Digital Berbasis EfficientNetV2 dengan Visualisasi Explainable AI. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002221063-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Praktik pemalsuan bubuk cabai merah yang merugikan konsumen saat ini masih sering diuji dengan metode konvensional yang bersifat destruktif dan memerlukan waktu yang lama. Penelitian ini mengusulkan penerapan metode berbasis pengolahan citra digital menggunakan model EfficientNetV2 untuk mengklasifikasikan citra bubuk cabai merah murni dan bubuk cabai merah yang telah dipalsukan, serta mengintegrasikan Explainable AI (XAI) melalui Grad-CAM untuk visualisasi area yang memengaruhi keputusan klasifikasi. Penelitian ini meggunakan data sekunder yang diperoleh dari website Mendeley Data sebanyak 4.916 citra yang mencakup 16 kelas, yang terdiri dari bubuk cabai murni serta lima jenis bahan pemalsu. Tahapan metodologi yang dilakukan mencakup pengumpulan data, pra-pemrosesan data, pembagian data, serta pelatihan dan pengujian model. Evaluasi kinerja model akan diukur menggunakan beberapa metrik, yaitu akurasi, presisi, recall, dan F1-score, serta analisis loss function menggunakan cross entropy loss. Dua skenario percobaan dilakukan dalam penelitian ini, yaitu penggunaan unsharp masking dan tanpa unsharp masking, yang dilakukan pada empat variasi fully connected layer. Hasil percobaan menunjukkan performa terbaik diperoleh pada model tanpa penggunaan unsharp masking dengan variasi 16 unit fully connected layer, yang mencapai nilai recall sebesar 0,9712.
===================================================================================================================================
The adulteration of red chili powder, which harms consumers, is still commonly tested using conventional methods that are destructive and time-consuming. This study proposes the application of a digital image processing-based method using the EfficientNetV2 model to classify images of pure red chili powder and adulterated red chili powder, as well as integrating Explainable AI (XAI) through Grad-CAM to visualize the areas influencing the classification decision. This study uses secondary data obtained from the Mendeley Data website, consisting of 4,916 images covering 16 classes, comprising pure chili powder and five types of adulterants. The methodological stages carried out include data collection, data pre-processing, data splitting, as well as model training and testing. Model performance will be evaluated using several metrics, namely accuracy, precision, recall, and F1-score, along with loss function analysis using cross entropy loss. Two experimental scenarios were conducted in this study, namely with and without unsharp masking, applied to four variations of fully connected layer configurations. The experimental results show that the best performance was obtained by the model without unsharp masking with a variation of 16 units in the fully connected layer, achieving a recall value of 0,9712.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Pemalsuan Bahan Pangan, Bubuk Cabai Merah, Klasifikasi, EfficientNetV2, XAI (Grad-CAM), Food Adulteration, Red Chili Powder, Classification, EfficientNetV2, XAI (Grad-CAM).
Subjects:	Q Science Q Science > QA Mathematics Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Mathematics and Science > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Nabilah Putri Srimaharani
Date Deposited:	27 Jul 2026 04:10
Last Modified:	27 Jul 2026 04:10
URI:	http://repository.its.ac.id/id/eprint/137458

Actions (login required)

View Item