Back To Projects
Back To Projects

Structural damage classification of civil structures for a global oil and gas conglomerate

SUPA's experts scaled client's data annotation, accurately annotating 12,000+ images to boost damage classification workflow.

Problem statement

The Client required a specialised team to identify and assess damage of civil structures with a high accuracy of at least 90%.

SOLUTION

SUPA’s engineering experts collaborated closely with the Client by co-creating the annotation workflow and assembling a team of annotators experienced with engineering-related projects.

RESULT

SUPA’s team of 25 annotators successfully delivered the annotations with a consistent >90% accuracy.

Overview

The Problem

The Client is an integrated Oil & gas company looking to build an automated damage classification model that could:

  1. Identify the structure of interest within an image.
  2. Detect any structural damage and classify the severity of the damage based on the inspection team’s internal best practices.

The Client does not have an internal labeling team. Additionally, their previous 3rd party vendor struggled with generating annotations with a minimum accuracy of 90%.

The Solution

SUPA recognized the importance of understanding the task’s engineering context and the open-endedness associated with assessing structural damage. The team:

  1. Collaborated with the client to co-create labeling instructions and workflow: Our project managers worked closely with the client’s civil engineers through Proof-of-Concepts to curate a scalable and comprehensive workflow.

  1. Deployed expert annotators: Only annotators with proven experience in engineering projects were assigned to this project to ensure consistent, high-quality annotations.

  1. Ran a continuous feedback loop: Project managers conducted weekly feedback meetings to address new edge cases and expand the team’s knowledge base.

The Result

SUPA successfully delivered all the data required by the Client, completing over 22,500 annotations at a consistent >90% accuracy.

This enabled the Client to enhance their classification model, enabling a faster, safer, and more accurate inspection process.

Why SUPA?

SUPA’s engineering experts go beyond generic annotation work. They understand the underlying context of the task, working closely with the Client from Day 1 to build a tailored work process to successfully deliver the annotation project.

Other Projects

Discover the work we do

View All
View All

Data labeling for autonomous vehicle training

Discover how SUPA's specialized data labeling services enhanced an autonomous driving company's models, achieving 95% accuracy

problem
The client needed to enhance their autonomous driving model's accuracy in interpreting vector spaces to meet safety standards and effectively navigate diverse environments.
solution
SUPA provided expert data labeling services, combining rigorous annotator training and human-machine collaboration, along with a dedicated quality control team to ensure high-quality, consistent data.
result
SUPA continues to deliver labeled data to the Client with 95% accuracy, significantly improving the client’s model predictions and meeting tight delivery timelines since 2022.
95% accuracy
Autonomous Vehicles
Computer Vision
Data labeling for autonomous vehicle training

SUPA’s strategic use of graphic design experts for generative model excellence

SUPA leveraged domain-specific talent to source and label 600k stylized vector images, solving data diversity challenges

problem
The client struggled to source and label 600,000 diverse vector images for training a stylistically versatile GenAI model, hampered by inconsistent vendor quality.
solution
SUPA rapidly scaled domain-expert graphic designers to curate, label, and sketch assets via iterative workflows and ML-aligned quality checks.
result
SUPA delivered the dataset in three months with only 3.5% rework, enabling a 37% FID score improvement in the client’s AI model for style-diverse outputs.
SUPA’s strategic use of graphic design experts for generative model excellence