Comparative Evaluation of TensorFlow Lite Quantization Variants for MobileNetV2-Based Batik Motif Classification: Accuracy, Model Size, and Inference Latency
DOI:
https://doi.org/10.30865/jurikom.v15i5.10047Keywords:
Batik Image, CNN, Post-Training Quantization, TensorFlow Lite, Transfer LearningAbstract
Convolutional Neural Network (CNN) models achieve high accuracy in visual recognition, but their size and computational cost limit deployment on mobile devices, so batik motif recognition remains difficult to bring to artisans and small creative enterprises. This study evaluates how post-training quantization with TensorFlow Lite affects the classification accuracy, model size, and inference latency of a MobileNetV2 transfer-learning model for batik motif recognition, in order to identify the deployment configuration that offers the best balance among the three. The experimental workflow adapted a pre-trained MobileNetV2 network to the Batik Wastra Nusantara dataset, which covers batik designs originating from different parts of Indonesia with substantial variation in color, ornamentation, and pattern complexity, converted the trained model into Dynamic Range, Float16, and Int8 TensorFlow Lite variants, and compared them under identical conditions. Evaluation on 644 valid images from 14 motif classes produced a test accuracy of 43.08%, corresponding to 28 of the 65 test images, a macro F1-score of 41.65%, and a weighted F1-score of 40.39%. Across the four model formats the accuracy spans only three test images, from 41.54% to 46.15%, a range that lies well inside the confidence interval implied by a 65-image test subset, so the variants cannot be separated on accuracy at this sample size. The deployment metrics differ substantially: relative to the unquantized reference, the Float16 variant reduced model size from 19.04 MB to 4.29 MB (77.47%) and inference latency from 205.52 ms to 28.60 ms (86.08%), while the Int8 variant was the fastest at 14.87 ms and the Dynamic Range variant the smallest at 2.41 MB. These findings indicate that the benefit of TensorFlow Lite optimization lies in storage and latency rather than in accuracy, and that the accuracy attainable on this dataset is constrained by its limited size and the visual similarity among motifs. The study contributes the deployment measurements, namely model size and inference latency for each quantization variant, that comparable batik classification studies do not report, and identifies dataset expansion as the prerequisite for reaching production-level accuracy in AI-based digital innovation for Indonesia’s creative industries.
References
[1] M. Aprianti, D. A. Dewi, and Y. F. Furnamasari, “Kebudayaan Indonesia di Era Globalisasi Terhadap Identitas Nasional Indonesia,” Edumaspul: Jurnal Pendidikan, vol. 6, no. 1, pp. 996–998, Mar. 2022, doi: 10.33487/edumaspul.v6i1.2294.
[2] A. D. W. Sumari, N. L. Aziza, and M. Hani’ah, “Preserving Indigenous Indonesian Batik Motif Using Machine Learning and Information Fusion,” JOIV : International Journal on Informatics Visualization, vol. 9, no. 5, p. 2018, Sep. 2025, doi: 10.62527/joiv.9.5.3714.
[3] A. Dhillon and G. K. Verma, “Convolutional neural network: a review of models, methodologies and applications to object detection,” Progress in Artificial Intelligence, vol. 9, no. 2, pp. 85–112, Jun. 2020, doi: 10.1007/s13748-019-00203-0.
[4] A. Voulodimos, N. Doulamis, A. Doulamis, and E. Protopapadakis, “Deep Learning for Computer Vision: A Brief Review,” Comput. Intell. Neurosci., vol. 2018, pp. 1–13, 2018, doi: 10.1155/2018/7068349.
[5] N. Sharma, V. Jain, and A. Mishra, “An Analysis Of Convolutional Neural Networks For Image Classification,” Procedia Comput. Sci., vol. 132, pp. 377–384, 2018, doi: 10.1016/j.procs.2018.05.198.
[6] S. Subramanian et al., “Mobile Object Detection using TensorFlow Lite and Transfer Learning,” in IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017.
[7] F. Windhager et al., “Visualization of Cultural Heritage Collection Data: State of the Art and Future Challenges,” IEEE Trans. Vis. Comput. Graph., vol. 25, no. 6, pp. 2311–2330, Jun. 2019, doi: 10.1109/TVCG.2018.2830759.
[8] D. Chen and P. Cheng, “A method to extract batik fabric pattern and elements,” Journal of the Textile Institute, vol. 112, no. 7, 2021, doi: 10.1080/00405000.2020.1802885.
[9] Z.-P. Jiang, Y.-Y. Liu, Z.-E. Shao, and K.-W. Huang, “An Improved VGG16 Model for Pneumonia Image Classification,” Applied Sciences, vol. 11, no. 23, p. 11185, Nov. 2021, doi: 10.3390/app112311185.
[10] Mehrunnisa, M. Leszczuk, D. Juszka, and Y. Zhang, “Improved Binary Classification of Underwater Images Using a Modified ResNet-18 Model,” Electronics (Basel)., vol. 14, no. 15, p. 2954, Jul. 2025, doi: 10.3390/electronics14152954.
[11] K. Zhao et al., “Efficient memory reuse methodology for CNN-based real-time image processing in mobile-embedded systems,” J. Real. Time. Image Process., vol. 20, no. 6, p. 118, Dec. 2023, doi: 10.1007/s11554-023-01375-8.
[12] 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,” Technologies (Basel)., vol. 12, no. 6, p. 81, Jun. 2024, doi: 10.3390/technologies12060081.
[13] D. G. T. Meranggi, N. Yudistira, and Y. A. Sari, “Batik Classification Using Convolutional Neural Network with Data Improvements,” International Journal on Informatics Visualization, vol. 6, no. 1, 2022, doi: 10.30630/joiv.6.1.716.
[14] H. Sastypratiwi, H. Muhardi, and Y. Yulianti, “Batik Recognition and Classification Using Transfer Learning and MobileNet Approach,” JOIV : International Journal on Informatics Visualization, vol. 8, no. 4, p. 2400, Dec. 2024, doi: 10.62527/joiv.8.4.2407.
