Created an app that generates realistic and customizable driver faces, providing a low-cost stream of high-impact training and validation data to improve the accuracy of autonomous-driving models.

Features

• Realistic Face Generation: Creates realistic driver faces using state-of-the-art ML models
• Customizable data: Allows control over facial features, expressions, and demographics
• High-Impact: Provides real-world scenarios that improve model accuracy
• Cost-Effective: Low-cost alternative to collecting real-world driver data
• Production-ready pipeline: Can be used to generate large training datasets

Impact

The solution enabled an autonomous-driving company to expand its training and validation data which improved its AD models. It reduced cost, and saved time by not having to collect diverse and rare data, without compromising on realism.