Modamia, a modern fashion retail brand, aimed to deliver hyper-personalized shopping experiences by enabling customers to design outfits suited to their style, body type, and occasion. But as their customer base grew, so did the challenges of offering visual recommendations at scale. Partnering with GoML, Modamia implemented a machine learning and AI fashion personalization engine powered by CNNs and body measurement estimation, creating a seamless and customized fashion experience.

The problem: poor visualization and lack of personalized fashion recommendations

Modamia’s core offering, customer-designed outfits, was unique but overwhelming for many users. Shoppers struggled to visualize how various clothing elements would look on their body, leading to friction in the buying journey.

Additionally, existing recommendation systems fell short in addressing individual style preferences, occasion-specific needs, and unique body shapes, resulting in generic suggestions and lower conversions.

The solution: A visual AI fashion personalization engine for outfits

GoML built a tailored AI pipeline to deliver personalized, photorealistic outfit suggestions and dynamic customization using deep learning and AWS-native services.

Photo-based outfit recommendations

Customers uploaded their photos and entered prompts like “I want a golden gown for a party.”

Using CNN models, the system generated visual representations of outfits directly on user images, powered by real-time data processing in MongoDB. This allowed customers to see themselves in the outfits before purchase.

Real-time dynamic adjustments

Users could further refine outfits by specifying style elements such as “make it sleeveless” or “add a slit”.

The ML system dynamically adapted the outfit in real time, providing updated, accurate visuals aligned with the user's exact preferences.

Personalized body shape visualization

To improve fit accuracy and realism, the system used CNNs (EfficientNet + MLP layers) to estimate a customer’s key body measurements from two photos (front and side) plus height and gender data.

Training data used: AWS Body Measurements Open Dataset

Front-end: Built using React.js for a seamless experience.

The impact: enhanced customer experience with AI fashion personalization

Modamia saw measurable success across engagement, conversions, and satisfaction:

• 85% increase in engagement, thanks to visual, personalized outfit previews

• 75% boost in conversion rates, driven by dynamic customization and better fit estimation

• 90% improvement in customer satisfaction, as shoppers could truly “see themselves” in their purchases

• 6.5 MAE on human body measurement prediction, ensuring realistic visual outputs

Lessons for other fashion retailers

Common pitfalls to avoid

• Using generic recommendation engines that ignore body type and occasion

• Skipping visualization, which reduces shopper confidence

• Not accounting for regional fashion styles and preferences

Advice for scaling fashion AI

• Integrate real-time customization for higher conversions

• Use body estimation to improve product fit and returns

• Combine prompt-based input with visual output to drive engagement

Want to offer AI-driven fashion personalization for your customers?

Let GoML help you bring style, accuracy, and experience with our AI engine and AI boilerplates for personalization.