POV Fit Engine — Don't Just Add It to Your Cart. Fit It with POV.

Inspiration

Online fashion shopping offers thousands of choices, but two important questions remain unanswered: Does this match my personal taste, and can I see myself wearing it?

Traditional recommendation systems often suggest products based on previous clicks or purchases. Yet fashion preferences are influenced by much more than shopping history. Music, designers, cultural trends, personal interests and individual expression all shape what people choose to wear.

That inspired POV Fit Engine, a personalised virtual fitting room designed to understand the shopper's point of view.

Our vision is to move beyond generic recommendations and bring together cultural taste intelligence, machine-learning-based preference prediction and AI-powered virtual try-on.

With Qloo, POV can explore relationships between a shopper's selected interests and relevant cultural tastes. With our XGBoost recommendation model, it can estimate the likelihood of a shopper liking a product. With Perfect Corp's YouCam Apparel Virtual Try-On, shoppers can visualise selected garments on their own photos.

The result is a shopping experience built around a simple idea:

Discover → Personalize → Fit → Decide → Buy.

What We Built

POV Fit Engine is a web-based fashion discovery and virtual try-on platform built with Python and Flask.

The experience brings together three complementary capabilities:

1. Machine-learning-powered recommendations

Our XGBoost model estimates a shopper's likelihood of liking a clothing item using the available user and product features. This helps rank products according to predicted preference rather than relying exclusively on a generic catalog.

2. Culture-aware personalisation with Qloo

The planned Qloo integration adds a taste-intelligence layer to product discovery. Using supported taste signals and relationships, POV can explore connections between a shopper's selected interests and fashion preferences.

Instead of treating culture as a demographic label, POV uses voluntarily supplied interests to help make discovery more relevant to the individual.

3. Virtual try-on with Perfect Corp YouCam

When a shopper finds something interesting, they can select Fit with POV to generate an AI-powered garment preview using their photo and the selected product's reference image.

This turns a recommendation into an interactive decision-making experience, helping shoppers visualise clothing before deciding whether to purchase it.

How We Built It

POV combines a Flask application, a machine-learning recommendation layer, product data and Perfect Corp's Apparel Virtual Try-On API.

Our architecture separates the key responsibilities:

• Flask manages the application and shopping experience.
• XGBoost provides product-preference predictions.
• Qloo Insights API adds cultural taste intelligence and related preference signals.
• Perfect Corp YouCam generates virtual garment previews.
• The product catalog connects recommendations and try-on requests to real product images and garment metadata.

The goal is to make POV's intelligence reusable beyond a standalone store. The personalisation and try-on capabilities could ultimately be exposed as services that online retailers integrate into their existing shopping experiences.

What Makes POV Different?

Most fashion recommendations answer, "What might you like?"

POV aims to answer two connected questions:

"What matches your taste?"

"Can you see yourself wearing it?"

By connecting taste intelligence, preference prediction and visualisation, POV brings product discovery closer to the point where a shopper makes a decision.

We also recognise that a generated preview is a visual aid, not a guarantee of physical garment fit. Our goal is to give shoppers additional information and confidence, not to replace their judgment.

Challenges

One technical challenge is ensuring that virtual try-on works across different garment categories. A shirt, a dress, a pair of trousers and a complete coordinated outfit do not necessarily use the same reference-image requirements or processing category.

We are addressing this by making garment selection category-aware and aligning the try-on request with the supported API configuration. Complete outfits require particular care because multiple garments may need to be processed in sequence rather than submitted as a single composite image.

Another challenge is making cultural taste intelligence useful without making the experience complicated. The integration must translate taste signals into relevant products while preserving user choice, transparency and control.

What We Learned

Personalisation becomes more valuable when it helps someone take a meaningful next step.

A recommendation can introduce a product. Cultural taste intelligence can help make that discovery more relevant. Virtual try-on can then help the shopper visualise the item before deciding what to do next.

The most important lesson is that AI should not simply be added to every stage of an application. It should be applied where it meaningfully improves the user's experience.

Why POV?

Fashion is a form of personal expression. People with different interests, influences and preferences can have entirely different ideas of what makes an outfit right for them.

POV Fit Engine aims to make online shopping feel less like browsing an endless catalog and more like having a personal fitting room that understands your point of view.

Don't just add it to your cart. Fit it with POV.

Built With

• flask
• python
• qloo
• sqlalchemy
• xgboost