A part of FEV Group
Analytics & AI: vital for ADAS/AD product development
Used solutions
-
Cloud & Data Solutions
Find the latest events here
A part of FEV Group
Introduction
At FEV.io, we design transformative solutions by integrating AI, engineering analytics, and cutting-edge technologies. Our Automated LIDAR, Image, and Video Annotator (ALiVA) framework revolutionizes data annotation processes, addressing the challenges of creating high-quality reference datasets for ADAS/AD development.
Background
We partnered with a global OEM to develop a comprehensive ALiVA Framework for automated data annotation. The project focused on streamlining the preparation of large-scale datasets for training ADAS/AD models, encompassing diverse tasks like sensor fusion, object tracking, and scenario detection.
Approach
1. Automated Data Annotation:
- Developed state-of-the-art AI models for vision and LIDAR data annotation, ensuring high precision and reliability.
- Implemented scenario and event detection algorithms for segmentation across diverse driving conditions.
- Designed cognitive algorithms to eliminate false annotations and optimize data quality.
2.Semi-Automated Manual Review:
- Introduced a manual review mechanism for error rectification, ensuring superior data validation.
- Enabled operational analytics reporting to track annotation times and identify improvement areas.
- Enhanced efficiency with a self-learning architecture for continuous optimization.
3. Customizable and Scalable Solutions:
- Tailored the annotation framework to meet client-specific requirements, ensuring relevance for various scenarios.
- Built a scalable system capable of handling on-road and off-highway applications for diverse industries.
4.Enhanced Engineering Analytics:
- Integrated interactive dashboards for real-time monitoring of annotation processes.
- Provided advanced analytics for object tracking, distance estimation, and cognitive filtering.
- Delivered actionable insights to improve the accuracy and efficiency of training datasets.
Impact
This innovative approach delivered measurable outcomes:
- Achieved over 80% savings in expert manual efforts, drastically reducing annotation time.
- Realized a 50% improvement in time-to-market for ADAS/AD model training datasets.
- Delivered precise, reliable, and scalable annotation solutions adaptable to multiple industries.
The ALiVA framework redefined the data annotation process, setting a benchmark for efficiency, accuracy, and cost-effectiveness in ADAS/AD product development. By leveraging AI and advanced analytics, it empowered the OEM to enhance its product development lifecycle significantly.
