SALSABILLA, SYLVIA THALIA (2025) KLASIFIKASI CITRA REMPAH – REMPAH INDONESIA BERBASIS TRANSFER LEARNING PROGRESIF PADA DENSENET-121. Skripsi thesis, UPN Veteran Yogyakarta.
![[thumbnail of 1_Skripsi Full_123210057_Sylvia Thalia Salsabilla.pdf]](http://eprints.upnyk.ac.id/style/images/fileicons/text.png) |
Text
1_Skripsi Full_123210057_Sylvia Thalia Salsabilla.pdf Restricted to Repository staff only Download (5MB) |
|
Text
2_Cover_123210057_Sylvia Thalia Salsabilla.pdf Download (160kB) |
|
Text
3_Abstrak_123210057_Sylvia Thalia Salsabilla.pdf Download (228kB) |
|
Text
4_Halaman Pengesahan_123210057_Sylvia Thalia Salsabilla.pdf Download (883kB) |
|
Text
5_Daftar Isi_123210057_Sylvia Thalia Salsabilla.pdf Download (213kB) |
|
Text
6_Daftar Pustaka_123210057_Sylvia Thalia Salsabilla.pdf Download (204kB) |
Abstract
ABSTRAK
Indonesia dikenal sebagai negara yang kaya akan rempah – rempah, yang tidak hanya
memiliki nilai ekonomi tinggi tetapi juga penting dalam berbagai industry seperti pangan,
farmasi, dan kosmetik. Namun, kemiripan bentuk dan warna dari beberapa jenis rempah
seringkali menyebablan kesalahan dalam proses identifikasi, yang berisiko menimbulkan
pemalsuan dan penipuan produk. Hal ini dapat berdampak pada kualitas bahan baku dan
kepercayaan konsumen terhadap produk rempah Indonesia. Oleh karena itu, dibutuhkan
upaya yang lebih dalam klasifikasi jenis rempah – rempah secara efektif dan akurat.
Penelitian ini menggunakan metode Convolutional Neural Network (CNN) dengan
arsitektur DenseNet121. Dataset yang digunakan terdiri dari 6.510 citra rempah – rempah
dengan 31 kelas. Penelitian ini mencakup tahap pengumpulan data, preprocessing (resize,
augmentasi, normalisasi), pembagian data latih dan data uji, serta pelatihan model dengan
lima variasi strategi transfer learning, mulai dari tanpa transfer learning hingga full transfer
learning. Evaluasi performa model dilakukan dengan menggunakan metrik accuracy,
precision, recall, dan F1 – Score untuk mengetahui efektivitas model dalam
mengklasifikasikan jenis rempah – rempah secara tepat.
Hasil penelitian menunjukkan bahwa model dengan full transfer learning yang
dimana seluruh lapisan DenseNet121 dilatih ulang, menghasilkan performa terbaik dengan
accuracy sebesar 90,71%, precision 91,27%, recall 90,71%, dan F1 – Score 90,65%.
Penelitian ini membuktikan bahwa semakin dalam proses fine – tuning dilakukan, semakin
baik pula kemampuan model dalam mengenali fitur yang spesifik dari masing – masing jenis
rempah.
Kata Kunci: Klasifikasi Citra, Convolutional Neural Network (CNN), DenseNet121,
Transfer Learning, Rempah – Rempah.
vii
ABSTRACT
Indonesia is known as a country rich in spices, which not only have high economic
value but are also important in various industries such as food, pharmaceuticals, and
cosmetics. However, the similarity in shape and color of several types of spices often causes
errors in the identification process, which risks product counterfeiting and fraud. This can
have an impact on the quality of raw materials and consumer confidence in Indonesian spice
products. Therefore, more efforts are needed in classifying types of spices effectively and
accurately.
This study uses the Concolutional Neural Network (CNN) method with the
DenseNet121 architecture. The dataset used consists of 6.510 spice images with 31 classes.
This study includes the stages of data collection, preprocessing (resize, augmentation,
normalization), division of training data and test data, and model training with five
varaitions of transfer learning strategies, ranging from no transfer learning to full transfer
learning. Model performance evaluation is carried out using accuracy, precision, recall, and
F1 – Score metrics to determine the effectiveness of the model in classifying spice types
accurately.
The results of the study showed that the model with full transfer learning where all
DenseNet121 layers were retrained, produced the best performance with an accuracy of
90,71%, precision of 91,27%, recall of 90,71%, and F1 – Score og 90,65%. This study proves
that the deeper the fine – tuning process is carried out, the better the model’s ability to
recognize specific features of each type of spice.
Keyword: Image Classification, Convolutional Neural Network (CNN), DenseNet121,
Transfer Learning, Spices.
viii
| Item Type: | Tugas Akhir (Skripsi) |
|---|---|
| Uncontrolled Keywords: | Klasifikasi Citra, Convolutional Neural Network (CNN), DenseNet121, Transfer Learning, Rempah – Rempah. |
| Subjek: | Z Bibliography. Library Science. Information Resources > ZA Information resources |
| Divisions: | Fakultas Teknik Industri > (S1) Informatika |
| Depositing User: | Eko Yuli |
| Date Deposited: | 14 Oct 2025 02:13 |
| Last Modified: | 14 Oct 2025 02:13 |
| URI: | http://eprints.upnyk.ac.id/id/eprint/44314 |
Actions (login required)
 |
View Item |