Analisis Kemampuan Beta-VAE Pada Dataset Yang Berbeda
Abstract
Abstrak - Data sintetis sudah menjadi beberapa penelitian untuk kasus machine learning, salah satunya adalah menambah data baru dikarenakan kurangnya data yang sudah ada. Tetapi bagaimana untuk menghasilkan dan mengatur berbagai variasi dari distribusi data masukan masih menjadi bahan penelitian. Pada penelitian ini menggunakan salah satu variasi metode Variational Auto Encoder (VAE) untuk menghasilkan data sintetis, yaitu Beta-Variational Auto Encoder (Beta-VAE). VAE sendiri merupakan metode unsupervised learning yang dapat menghasilkan data sintetis, tetapi variasi yang dihasilkan tidak terlalu teratur dibandingkan Beta-VAE. Pada penelitian ini digunakan metode Beta- VAE asli untuk menghasilkan data sintetis yang dilatih dengan empat dataset yang berbeda. Digunakan metrik PSNR, SSIM dan FID score untuk mengevaluasi model Beta-VAE. Dibandingkan setiap model Beta-VAE yang dilatih dengan dataset berbeda dan dilakukan analisis pada setiap model. Hasil dari penelitian didapati model yang dilatih dengan CelebA memiliki hasil terbaik terlihat dari metrik evaluasi.
Kata kunci : data sintetis, dataset, autoencoder, variational autoencoder
References
Goodfellow, I. J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., … Bengio, Y. 2014. Generative Adversarial Networks.
Chen, X., Duan, Y., Houthooft, R., Schulman, J., Sutskever, I., & Abbeel, P. 2016. InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets. CoRR, abs/1606.03657. Retrieved from http://arxiv.org/abs/1606.03657
Kulkarni, T. D., Whitney, W., Kohli, P., & Tenenbaum, J. B. 2015. Deep Convolutional Inverse Graphics Network. CoRR, abs/1503.03167. Retrieved from http://arxiv.org/abs/1503.03167
Kingma, D. P., & Welling, M. 2022. AutoEncoding Variational Bayes.
Zhou, L., Deng, W., & Wu, X. 2020. Unsupervised anomaly localization using VAE and beta-VAE. CoRR, abs/2005.10686. Retrieved from https://arxiv.org/abs/2005.10686
Pandey, K., Mukherjee, A., Rai, P., & Kumar, A. 2021. VAEs meet Diffusion Models: Efficient and High-Fidelity Generation. In NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications. Retrieved from https://openreview.net/forum?id=-J8dM4ed_92
Higgins, I., Matthey, L., Pal, A., Burgess, C., Glorot, X., Botvinick, M., … Lerchner, A. 2017. beta- VAE: Learning Basic Visual Concepts with a Constrained Variational Framework. In International Conference on Learning Representations. Retrieved from https://openreview.net/forum?id=Sy2fzU9gl
Fil, M., Mesinovic, M., Morris, M., & Wildberger, J. 2021. Beta-VAE Reproducibility: Challenges and Extensions. CoRR, abs/2112.14278. Retrieved from https://arxiv.org/abs/2112.14278
Ihsan, A. F. 2023. Initial Study of Batik Generation using Variational Autoencoder. Procedia Computer Science, 227, 7853794. https://doi.org/https://doi.org/10.1016/j.procs.2023 .10.584
Liu, Z., Luo, P., Wang, X., & Tang, X. 2015. Deep Learning Face Attributes in the Wild. In Proceedings of International Conference on Computer Vision (ICCV).
Aubry, M., Maturana, D., Efros Alexei, Russell, B., & Sivic, J. 2014. Seeing 3D chairs: exemplar part- based 2D-3D alignment using a large dataset of CAD models. In CVPR.
Paysan, P., Knothe, R., Amberg, B., Romdhani, S., & Vetter, T. 2009. A 3D Face Model for Pose and Illumination Invariant Face Recognition. In 2009 Sixth IEEE International Conference on Advanced Video and Signal Based Surveillance (pp. 2963 301). https://doi.org/10.1109/AVSS.2009.58
Bank, D., Koenigstein, N., & Giryes, R. 2020. Autoencoders. CoRR, abs/2003.05991. Retrieved from https://arxiv.org/abs/2003.05991
Burgess, C. P., Higgins, I., Pal, A., Matthey, L., Watters, N., Desjardins, G., & Lerchner, A. 2018. Understanding disentangling in β-VAE. ArXiv, abs/1804.03599. Retrieved from https://api.semanticscholar.org/CorpusID:4879659
Fardo, F. A., Conforto, V. H., de Oliveira, F. C., & Rodrigues, P. S. S. 2016. A Formal Evaluation of PSNR as Quality Measurement Parameter for Image Segmentation Algorithms. CoRR, abs/1605.07116. Retrieved from http://arxiv.org/abs/1605.07116
Sara, U., Akter, M., & Uddin, M. S. 2019. Image Quality Assessment through FSIM, SSIM, MSE and PSNR4A Comparative Study. Journal of Computer and Communications, 07, 8318. https://doi.org/10.4236/jcc.2019.73002
Nilsson, J., & Akenine-Möller, T. 2020. Understanding SSIM. ArXiv, abs/2006.13846. Retrieved from https://api.semanticscholar.org/CorpusID:2200416 31
Wang, Z., Bovik, A., Sheikh, H., & Simoncelli, E. 2004. Image Quality Assessment: From Error Visibility to Structural Similarity. Image Processing, IEEE Transactions on, 13, 6003612. https://doi.org/10.1109/TIP.2003.819861
Lucic, M., Kurach, K., Michalski, M., Gelly, S., & Bousquet, O. 2017. Are GANs Created Equal? A Large-Scale Study. In Neural Information Processing Systems. Retrieved from https://api.semanticscholar.org/CorpusID:4053393
Subramanian, A. K. 2020. PyTorch-VAE. GitHub repository. GitHub.
Krause, J., Stark, M., Deng, J., & Fei-Fei, L. 2013. 3D Object Representations for Fine-Grained Categorization. In 2013 IEEE International Conference on Computer Vision Workshops (pp. 5543561). https://doi.org/10.1109/ICCVW.2013.77
Spandan Ghosh. (n.d.). Cats faces 64x64 (For generative models). Retrieved September 8, 2023, from https://www.kaggle.com/datasets/spandan2/catsfaces-64x64-for-generative-models/data
Dionisius Darryl Hermansyah. (n.d.). Indonesian Batik Motifs. Retrieved November 20, 2023, from https://www.kaggle.com/datasets/dionisiusdh/indo nesian-batik-motifs?select=batik-bali
He, K., Zhang, X., Ren, S., & Sun, J. 2015. Deep Residual Learning for Image Recognition. CoRR, abs/1512.03385. Retrieved from http://arxiv.org/abs/1512.03385
