It uses the Bing Image Search API (part of Azure Cognitive Services) because it is simple, well‑documented, and returns clean JSON with the image URLs you need. If you prefer Google Custom Search, the code can be swapped with only a few lines – I’ll include a short note at the end.
5. Discussion
4. Results
5.1 The Jilbab as a Visual Mediator of Modern Modesty
The clustering results reveal a garment that is no longer monolithic. Instead, the jilbab functions as a visual canvas onto which global fashion vocabularies are projected. This mirrors the broader modest‑fashion narrative: a negotiation between religious/ cultural authenticity and contemporary aesthetic desire (Mahmood, 2020).
3.2 Automated Image Analysis
- Pre‑processing: Resizing to 224 × 224 px, colour normalisation.
- Feature Extraction: Pre‑trained ResNet‑50 model (He et al., 2016) fine‑tuned on a modest‑fashion subset (5 000 labelled images).
- Clustering: t‑SNE visualisation followed by k‑means (k = 8) to reveal dominant visual families.
