Key Takeaways
- • SunTec India achieved and sustained 99% accuracy across 2 million+ automotive and MRO (Maintenance, Repair, Operations) records processed for a global engineering conglomerate, demonstrating scalability of human-in-the-loop quality assurance workflows.
- • The company scaled annotation operations from 4 annotators and 1 reviewer to 35 annotators and 7 reviewers while maintaining accuracy benchmarks, crediting success to deliberate reviewer-to-annotator ratios and structured subject-matter expert training aligned with IRC82 government standards.
- • Completed a major data labeling initiative covering 1,000+ kilometers of highway infrastructure across India's national and state networks, with 71 distinct damage and asset categories classified using bounding box and multi-point polygon annotation techniques.
- • Rohit Bhateja, Director of Digital Engineering Services and Head of Marketing at SunTec India, emphasized the operational rigor required to maintain accuracy across edge cases introduced by regional variations in weather, lighting, and road conditions.
Summary
SunTec India completed a significant data annotation project for a government-backed infrastructure consulting firm, processing more than 3 million annotations with sustained 99% accuracy across automotive and maintenance records. The engagement involved labeling highway images to train a damage detection model for infrastructure monitoring across India's national and state highway networks, covering over 1,000 kilometers of corridor and encompassing 71 distinct damage and asset categories in compliance with IRC82 government standards.
The project's complexity reflected the demanding nature of highway asset classification. Approximately 80% of the workload required 4-point polygon annotations, while the remaining 20% demanded highly detailed multi-point annotations to capture nuanced damage patterns and asset distinctions. Operating on a client-hosted CVAT platform, SunTec India's team applied precision annotation methodologies to standardize labeling across diverse regional conditions where weather, lighting, and wear patterns vary significantly.
Operational scalability proved critical to project success. The initial pilot phase employed 4 annotators and 1 reviewer. Following successful validation, the scope expanded eightfold, growing to 35 annotators and 7 reviewers—all subject-matter experts with civil engineering backgrounds. The company maintained its 99% accuracy benchmark through this dramatic scaling by maintaining a disciplined reviewer-to-annotator ratio and implementing structured training protocols on IRC82 standards before live data labeling commenced.
Rohit Bhateja, Director of Digital Engineering Services at SunTec India, highlighted the significance of maintaining accuracy across the dataset's inherent complexity. He noted that the sustained accuracy rate across varied regional conditions and thousands of edge cases demonstrates the robustness of SunTec's human-in-the-loop workflow, even when dataset scope and annotator team size expand materially. This result underscores SunTec India's positioning as a capable vendor for large-scale, standards-compliant data annotation work serving government-backed and enterprise infrastructure initiatives.
The project's complexity reflected the demanding nature of highway asset classification. Approximately 80% of the workload required 4-point polygon annotations, while the remaining 20% demanded highly detailed multi-point annotations to capture nuanced damage patterns and asset distinctions. Operating on a client-hosted CVAT platform, SunTec India's team applied precision annotation methodologies to standardize labeling across diverse regional conditions where weather, lighting, and wear patterns vary significantly.
Operational scalability proved critical to project success. The initial pilot phase employed 4 annotators and 1 reviewer. Following successful validation, the scope expanded eightfold, growing to 35 annotators and 7 reviewers—all subject-matter experts with civil engineering backgrounds. The company maintained its 99% accuracy benchmark through this dramatic scaling by maintaining a disciplined reviewer-to-annotator ratio and implementing structured training protocols on IRC82 standards before live data labeling commenced.
Rohit Bhateja, Director of Digital Engineering Services at SunTec India, highlighted the significance of maintaining accuracy across the dataset's inherent complexity. He noted that the sustained accuracy rate across varied regional conditions and thousands of edge cases demonstrates the robustness of SunTec's human-in-the-loop workflow, even when dataset scope and annotator team size expand materially. This result underscores SunTec India's positioning as a capable vendor for large-scale, standards-compliant data annotation work serving government-backed and enterprise infrastructure initiatives.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.