Fine-Tuning the Zhang et al. Image Colorization Model on Indian Cultural Heritage Domains

This paper presented a fully reproducible pipeline for fine-tuning the Zhang et al. image colorization model on Indian cultural heritage im-agery, implemented entirely within Google Colab’s free T4GPU envi-ronment. A seven-domain dataset of 1,233 images was curated, split 80/10/10, and augmented for underrepresented categories. A two-stage fine-tuning strategy produced a model achieving an overall PSNR of 20.35dB and SSIM of 0.8587 on a 107-image test set. Music and architecture domains achieved the strongest results (22.91dB and 21.94dB respectively), while festival imagery proved most challenging (18.02dB) due to extreme palette diversity. The complete six-cell Colab workflow is described in sufficient detail for full reproducibility without dedicated GPU hardware, making this approach accessible to resource-constrained researchers in cultural heritage computing.

[1]     Zhang, R., Isola, P., & Efros, A.A. (2016). Colorful Image Colorization. European Conference on Computer Vision (ECCV) 2016. https://arxiv.org/abs/1603.08511 [2]     He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep Residual Learning for Image Recognition. CVPR2016. [3]     Goodfellow, I., Bengio, Y., & Courville, A. (2016).Deep Learning. MIT Press. [4]     Pedregosa, F., et al. (2011). Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 12, 2825–2830. https://scikit-learn.org/ [5]     Iizuka, S., Simo-Serra, E., & Ishikawa, H. (2016). Let there be color! Joint end-to-end learning of global and local image priors for automatic image colorization with simultaneous classification. ACM Transactions on Graphics (TOG), 35(4), 1–11. [6]     Su, J., Chu, H., & Jia, J. (2020).Instance-aware Image Colorization. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2020. [7]     Isola, P., Zhu, J.Y., Zhou, T., & Efros, A.A. (2017). Image-to-image translation with conditional adversarial networks.CVPR2017. [8]     Loshchilov, I., & Hutter, F. (2017).SGDR: Stochastic Gradient Descent with Warm Restarts. ICLR2017. [9]     Vitoria, P., Raad, L., & Ballester, C. (2020). ChromaGAN: Adversarial picture colorization with semantic class distribution. IEEE Winter Conference on Applications of Computer Vision (WACV) 2020. [10]     Wang, Z., Bovik, A.C., Sheikh, H.R., & Simoncelli, E.P. (2004). Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing, 13(4), 600–612.