Building an AI Model to Improve the Performance of Factory Workers

  • Client

    Industry: Research Services
    Location: Germany
    Size: 51–200 employees

  • Company Bio

    The client is a non-university research company with locations in Germany and Austria. The company conducts applied production research with a focus on industrial assistance systems and additive micro/nano manufacturing.

Services Provided

Image Annotation, Bounding Boxes, Keypoint Annotation

Project Overview

Mindy Support played a pivotal role in assisting the client to significantly enhance the precision and effectiveness of their AI model by delivering exceptional bounding box and keypoint annotation services, resulting in a remarkable improvement in overall performance.

Business Problem

Our client was working on an AI model that could monitor employees working at a factory. They had a sizable dataset of 2,000 images that needed to be annotated with bounding boxes and keypoint annotation. In addition to this, the client requested that all data annotators working on the project be located within the EU. All the work needed to be completed within a strict deadline of 1.5 months to avoid any costly delays. 

Why Mindy Support

The customer was considering multiple vendors and invited each company to complete a pilot project. Mindy Support was able to complete the pilot project with an exceptionally high accuracy rate and without any issues, which was exactly what the client was looking for. Our extensive experience actualizing similar image annotation projects further helped solidify our case for the client.

Solutions Delivered to the Client

As mentioned earlier, the client requested that all team members be located within the EU timezone, which meant that we needed to ramp up the team quickly to meet the deadline. We assembled a team of 12 data annotators, all located in EU countries, and had experience with our data annotation tool, CVAT. 

We decided to start by annotating cropped images with keypoint annotation, so we do not have to search for a worker in the image and zoom in. This output was sent to the client, so they could work with the data and further provide truncated images needing only bounding box annotations. In the event there were cropped images with two people, we discussed with the client that only the first person in the front needed to be annotated. 

Our team was able to annotate the entire dataset within the specified time period of 1.5 months with a quality rating of 99%+. The client was very satisfied with the results and recommended Mindy Support to their colleagues in the industry, and we are actively working to start actualizing those projects as well.

Key Results

  • 2,000 images annotated with bounding box and keypoint annotation
  • All the work was completed within 1.5 months
  • Ramped up the annotation team to 12 full-time annotators
  • 99%+ quality score

Get a quote for your data labeling project