Streamlining 3D Animation Data Labeling with Sapien

Building Precision: How Uthana Transformed Data Labeling

Uthana, a leader in computer vision solutions, encountered complex optimization challenges while managing a large-scale 3D animation labeling project.  To tackle the scale and scope of the project, they partnered with Sapien strategists to refine workflows and maintain accuracy, consistency and efficiency in their data labeling process. This collaboration not only improved the Uthana model performance, but it also laid the groundwork for sustained future success.

‍

“The Sapien team was helpful in chatting through our labeling needs and coming up with a solution that worked for us. They were flexible enough to work with our tools as well.” Viren Tellis, co-founder of Uthana.

Challenge: Tackling the Complexity of Large-Scale Labeling

Labeling over 10,000 data points for 3D animations demanded precision, consistency, and timely execution. Uthana’s internal workflows struggled with inefficiencies, and existing tools proved difficult to scale. These roadblocks delayed model training and impacted overall project timelines.

Solution: Tailored Workflows for Better Accuracy and Efficiency

Understanding the Problem

‍Sapien started with a full review of Uthana’s labeling pipeline and tool stack. We collaborated closely to map out what was slowing the process down and where improvements could be made.

Segmenting the Work, Step by Step‍

Instead of assigning complex, multi-step tasks to each labeler, Sapien helped break down the workflow into clearly defined task stages. Each stage was simpler, easier to train for, and more measurable. This allowed us to:

Rolling It Out in Phases

The updated workflow was introduced step-by-step, allowing both teams to review and adjust as we went. With regular feedback and open communication, we stayed aligned and made sure everything was working smoothly.

Results: Meeting and Exceeding Expectations

Project Highlights:

‍Project Improvements: Our structured and consistent labeling process enabled Uthana to integrate results seamlessly into their AI model training. The collaboration improved workflow efficiency and ensured high-quality data, resolving previous challenges and setting a foundation for future scalability.

Achieving More AI Innovation Together

This partnership between Uthana and Sapien demonstrates the power of collaboration. By creating custom solutions and fostering open communication, Sapien helped Uthana overcome complex challenges in data labeling.

476% Increase in Labeling Efficiency
Labeling throughput jumped from 20 labels per day to 115, drastically cutting down on time needed to complete tasks. The faster workflow helped Uthana stay on track without compromising on quality.

16% Improvement in Labeling Accuracy
By focusing training on individual steps, Uthana improved internal accuracy from 82% to 98%. Higher accuracy meant better training data and better-performing AI models.

Over 2,757 Minutes of Labeled Data Processed
This large-scale effort provided high-quality data for Uthana’s model development and future improvements.

Clear Results from a Simple Shift

Uthana didn’t just need labelers—they needed a better way to label. Sapien delivered a clear, practical workflow redesign that made the work easier, faster, and more accurate. If you’re facing similar scaling challenges in your AI labeling pipeline, Sapien can help you build workflows that work. Let’s talk.