Komparasi EasyOCR dan Tesseract dengan Klasifikasi SVM, k-NN, dan BiLSTM untuk Deteksi Kadar Natrium pada Label Gizi Produk Makanan dan Minuman Minimarket Berbasis FPN dan Faster R-CNN
Keywords:
Faster R-CNN, EasyOCR, Support Vector Machine, Class Imbalance, k-NN, Natrium TersembunyiAbstract
Deteksi informasi gizi dari gambar kemasan produk untuk pemantauan kesehatan secara instan menghadapi dua tantangan utama: variasi ukuran dan posisi teks pada gambar kemasan, serta minimnya data dan ketimpangan kelas yang ekstrem pada teks hasil ekstraksinya. Penelitian ini penting karena konsumsi natrium tersembunyi pada makanan olahan menjadi pemicu risiko penyakit kardiovaskular dan hipertensi, sehingga dibutuhkan sistem deteksi dini yang bisa diakses langsung oleh konsumen. Untuk itu, penelitian ini mengusulkan sebuah pipeline bernama SnackGuard yang menggabungkan Deep Learning di tahap awal (deteksi gambar) dengan model Machine Learning klasik di tahap akhir (klasifikasi kadar natrium). Di tahap awal, lokasi Informasi Nilai Gizi pada kemasan dideteksi menggunakan Faster R-CNN dan Feature Pyramid Network (FPN), lalu teksnya diekstrak dan dibandingkan menggunakan dua mesin OCR, yaitu Tesseract dan EasyOCR. Untuk mengatasi masalah kesalahan pembacaan angka dan kegagalan model dalam mempelajari teks pendek, string hasil OCR dibersihkan menggunakan ekspresi reguler (Regex) dan diubah menjadi satu nilai angka tunggal (skalar). Hasil pengujian menunjukkan bahwa model BiLSTM standar gagal mengenali kelas minoritas karena distribusi data yang timpang (kelas "Tinggi" hanya sekitar 4,5% dari total data). Sebaliknya, model k-Nearest Neighbor (k-NN) berhasil mencapai akurasi 99,17% dengan waktu proses yang sangat cepat (kompleksitas logaritmik), sementara Support Vector Machine (SVM) dengan kernel Linear juga terbukti tangguh dalam mengunci batas keputusan sesuai ambang gizi BPOM, dengan waktu proses yang konstan dan tanpa terpengaruh ketimpangan kelas. Penelitian ini membuktikan bahwa dengan menyederhanakan data teks menjadi satu nilai angka, kesalahan yang muncul dari tahap awal (OCR) bisa diminimalkan, sehingga pendekatan ini jauh lebih ringan, akurat, dan layak diterapkan pada perangkat dengan sumber daya terbatas dibandingkan model deep learning yang lebih kompleks
References
[1] Z. Guo, H. Shuai, G. Liu, Y. Zhu, and W. Wang, “Multi-level feature fusion pyramid network for object detection,” Visual Computer, vol. 39, no. 9, pp. 4267–4277, Sep. 2023, doi: 10.1007/s00371-022-02589-w.
[2] M. Hussain, W. Sharif, M. R. Faheem, Y. Alsarhan, and H. A. Elsalamony, “Cross-Platform Hate Speech Detection Using an Attention-Enhanced BiLSTM Model,” Engineering, Technology and Applied Science Research, vol. 15, no. 6, pp. 29779–29786, Dec. 2025, doi: 10.48084/etasr.13249.
[3] M. Nashiruddin, F. Dhaza Noor Praditya, and A. Faiz Mufazzal, “Implementasi OCR berbasis Tesseract untuk Ekstraksi data kartu mahasiswa UMKLA,” 2025.
[4] Z. Su, “Comparative Analysis of CNN-Based Object Detection Models: Faster R-CNN, SSD, and YOLO,” 2025.
[5] M. N. Darpito, K. Firdausy, and A. Fadlil, “JIP (Jurnal Informatika Polinema) Perbandingan Unjuk Kerja Library Optical Character Recognition (OCR) dalam Pengenalan Teks pada Dokumen Digital,” 2025.
[6] P. Sisodia and S. W. A. Rizvi, “Optical Charac-ter Recognition Development Using Python,” Journal of Informatics Electrical and Electronics Engineer-ing, vol. 04, no. 03, pp. 1–13, 2023, doi: 10.54060/jieee.202.
[7] M. Dzikry Pamungkas, N. Falih, M. Panji Muslim, and N. Nasrulloh, “OCR and NLP for Consumer Product Label Analysis: A Systematic Literature Review,” 2026. [Online]. Available: http://jurnal.polibatam.ac.id/index.php/JAIC
[8] K. A. Puteri, F. Alifia, P. A. Lailatulrahmi, G. P. Mindara, and E. P. Giri, “PERBANDINGAN AKURASI TESSERACT DAN EASYOCR SEBELUM DAN SESUDAH PRAPEMROSESAN PADA CITRA NOTA,” Jurnal Informatika dan Teknik Elektro Terapan, vol. 14, no. 1, Jan. 2026, doi: 10.23960/jitet.v14i1.8603.
[9] A. Hanif, A. Sahid, A. F. Anita, P. Studi Teknologi Informasi, F. Kesehatan dan Teknologi, and U. Muhammadiyah Klaten, “Perbandingan Kinerja EasyOCR dan Tesseract dalam Ekstraksi Teks pada citra Digital,” 2026.
[10] M. Ramzan, “A Deep CNN-BiLSTM Framework for Arabic Handwritten Text Recognition,” Engineering, Technology and Applied Science Research, vol. 16, no. 2, pp. 34171–34177, Jan. 2026, doi: 10.48084/etasr.17092.
[11] M. Qhitfir, H. Sujaini, and Y. Sholva, “Scientica,” 2025.
[12] B. Yu et al., “Deep Learning or Classical Machine Learning? An Empirical Study on Log-Based Anomaly Detection,” in Proceedings - International Conference on Software Engineering, IEEE Computer Society, May 2024. doi: 10.1145/3597503.3623308.
[13] R. Febrianti Malelak, A. Fanggidae, T. Widiastuti, and Y. T. Polly, “Informasi Kadar Natrium pada Label Produk Kemasan Menggunakan FPN, Faster R-CNN,” 2025.
[14] N. M. Min and M. M. Khaing, “MULTI-LABEL FOOD IMAGE DETECTION AND CLASSIFICATION WITH FASTER R-CNN,” International Research Journal of Modernization in Engineering Technology and Science, May 2025, doi: 10.56726/irjmets74972.
[15] Y. Shah, N. Jariwala, B. Kachhia, and P. Shah, “Delving Deep into NutriScan: Automated Nutrition Table Extraction and Ingredient Recognition,” 2023. [Online]. Available: www.ijraset.com
[16] V. J, D. Rajgopal, P. Yesubabu, and G. S, “Optical Character Recognition using Machine Learning and Deep Learning,” International Journal of Drug Delivery Technology, vol. 16, no. 49s, Jun. 2026, doi: 10.25258/ijddt.16.49s.131.
[17] T. M. E. Saputra, A. Susanto, and B. J. Carmelita, “Implementation of Tesseract OCR and Bounding Box for Text Extraction on Food Nutrition Labels,” Building of Informatics, Technology and Science (BITS), vol. 6, no. 3, pp. 1403–1412, Dec. 2024, doi: 10.47065/bits.v6i3.6107.
[18] O. Jouini, K. Sethom, A. Namoun, N. Aljohani, M. H. Alanazi, and M. N. Alanazi, “A Survey of Machine Learning in Edge Computing: Techniques, Frameworks, Applications, Issues, and Research Directions,” Jun. 01, 2024, Multidisciplinary Digital Publishing Institute (MDPI). doi: 10.3390/technologies12060081.
[19] L. Grinsztajn, E. Oyallon, and G. Varoquaux, “Why do tree-based models still outperform deep learning on typical tabular data?,” 2022.
[20] M. Nagayi, A. Khan, T. Frank, R. Swart, and C. Nyirenda, “Evaluating OCR performance on food packaging labels in South Africa,” Oct. 2025, [Online]. Available: http://arxiv.org/abs/2510.03570
[21] J. Yao and B. Yuan, “Optimization Strategies for Deep Learning Models in Natural Language Processing,” Journal of Theory and Practice of Engineering Science, vol. 4, no. 05, pp. 80–87, May 2024, doi: 10.53469/jtpes.2024.04(05).11.
[22] X. Xu et al., “Crack Detection and Comparison Study Based on Faster R-CNN and Mask R-CNN,” Sensors, vol. 22, no. 3, Feb. 2022, doi: 10.3390/s22031215.
[23] H. Yazdanjouei, A. Mansouri, and M. Shokouhifar, “A Co-Training Semi-Supervised Framework Using Faster R-CNN and YOLO Networks for Object Detection in Densely Packed Retail Images Email addresses,” 2025.
[24] S. U. Hassan, J. Ahamed, and K. Ahmad, “Analytics of machine learning-based algorithms for text classification,” Sustainable Operations and Computers, vol. 3, pp. 238–248, Jan. 2022, doi: 10.1016/j.susoc.2022.03.001.
[25] Z. Alshingiti, R. Alaqel, J. Al-Muhtadi, Q. E. U. Haq, K. Saleem, and M. H. Faheem, “A Deep Learning-Based Phishing Detection System Using CNN, LSTM, and LSTM-CNN,” Electronics (Switzerland), vol. 12, no. 1, Jan. 2023, doi: 10.3390/electronics12010232.
[26] M. Rafi-Ur-Rashid, M. Mahbub, and M. A. Adnan, “Breaking the Curse of Class Imbalance: Bangla Text Classification,” ACM Transactions on Asian and Low-Resource Language Information Processing, vol. 21, no. 5, Apr. 2022, doi: 10.1145/3511601.
[27] R. Anadra, H. Wijayanto, and K. Sadik, “Sentiment Analysis of Tokopedia Customer Reviews Using BiLSTM and IndoBERT with Comparative Analysis of Preprocessing and Labeling Methods,” International Journal of Advances in Data and Information Systems, vol. 6, no. 3, pp. 773–788, Dec. 2025, doi: 10.59395/ijadis.v6i3.1458.
[28] S. Dutta, S. Adhikary, and A. D. Dwivedi, “VisFormers—Combining Vision and Transformers for Enhanced Complex Document Classification,” Mach. Learn. Knowl. Extr., vol. 6, no. 1, pp. 448–463, Mar. 2024, doi: 10.3390/make6010023.
[29] A. N. Chowdhury et al., “Performance Analysis of Tesseract and EasyOCR for Bangla Optical Character Recognition on the Novel Bangla CrossHair Dataset,” 2025.
[30] C. Li, Z. Xie, and H. Wang, “Short Text Classification Based on Enhanced Word Embedding and Hybrid Neural Networks,” Applied Sciences (Switzerland), vol. 15, no. 9, May 2025, doi: 10.3390/app15095102.
Downloads
Published
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
