A naukri.com initiative
Swiggy
3w
31
Image Credit: Swiggy
Enhancing Brand Visibility and Trust with On device ML models: A Journey at Swiggy
- At Swiggy, trust is crucial for customer loyalty and brand success, focusing on creating seamless experiences from order to delivery.
- Enhancing brand visibility through branded gear for delivery partners was a key strategy to reinforce trust and professionalism.
- Using on-device machine learning (ML) models, Swiggy tackled the challenge of detecting Swiggy-branded gear on delivery executives in real-time.
- Server-side solutions posed challenges like high latency and real-time constraints, leading Swiggy to explore on-device solutions for faster feedback.
- Initially, a color detector approach was taken, but later, leveraging TensorFlow Lite Model Maker proved to be a more scalable and efficient solution.
- Training a MobileNet model and optimizing it for edge devices helped in achieving real-time gear detection with high accuracy and low latency.
- The on-device model had a latency of around 150ms, seamless integration with Vision Camera library, and optimizations for battery and CPU efficiency.
- The production rollout included system implementations for compliance checks and data gathering, leading to widespread adoption and positive impacts on latency and stability.
- Key takeaways included the power of edge ML for real-time solutions, the importance of optimization for performance enhancements, and the potential of on-device models for various use cases.
- Overall, integrating on-device ML models enhanced brand visibility, professionalism, and trust at Swiggy, ensuring a smoother experience for delivery partners and customers.
Read Full Article
Like
For uninterrupted reading, download the app