Adapting a Large Diffusion Transformer with Conditional Flow Matching for the Synthesis of Realistic Sri Lankan Faces

Contemporary text-to-image diffusion models, including FLUX.1-Dev and Stable Diffusion, exhibit representational bias stemming from their reliance on large-scale datasets that are potentially skewed towards certain ethnic distributions or rather lack of such diverse data in the training datasets. This imbalance results in systematic misrepresentation of underrepresented populations, such as Sri Lankans, in synthesised imagery. To address this gap, we construct a culturally curated dataset comprising 200 high-resolution facial images of Sri Lankan individuals, capturing intra-ethnic diversity across morphological features, skin tones, and traditional attire. We then perform full-parameter fine-tuning of the FLUX.1-Dev architecture-an advanced diffusion transformer model-on this dataset, employing conditional flow-matching objectives and a high-fidelity training pipeline optimised for cultural fidelity. Quantitative and qualitative evaluations reveal that the fine-tuned model exhibits markedly improved performance in generating faces consistent with Sri Lankan ethnomorphological characteristics. As part of this work, a human perceptual study (n <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$=134)$</tex> was conducted to assess the realism and cultural alignment of generated images. Statistical analysis via the Mann-Whitney U test confirms a significant improvement in perceived authenticity (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{p}&lt;10^{-48}$</tex>) over the baseline model. These results show marked efficacy of culturally grounded fine-tuning for mitigating demographic bias in generative vision models and highlight the necessity of domain-specific data for equitable machine learning pipelines.