Personalized outfit generation supported by machine learning
Pronti AI is transforming how consumers shop by connecting their existing closet data to purchasing decisions, supported by machine learning models.
Pronti AI wanted to enhance their existing clothing recommendation model to serve more relevant recommendations to their customers.
No automated system to identify clothing worn by models.
Trained and served an object detection model using bounding boxes to accurately detect items of clothing within retailers’ catalogue images.
Lack of data pipeline infrastructure for reliable job handling.
Implemented an airflow pipeline to ensure all aspects of image processing occur regularly with automatic retry, error handling, logging, and alerts.
No database to track the status of processed catalogue items.
Leveraged Google BigQuery to monitor item processing and state management.
Results
The Process
Discovery
Build Out
Outcomes
Outcomes
What was built
Object detection model with >95% accuracy
Model deployed on Cloud Run for scalable model serving
Composer model for automatic orchestration of image processing