SuperAnnotate

What do you like best?

SuperAnnotate took great care during the pilot process to ensure annotations met our needs. They extended our trial until they had an understanding of what features were essential and in scope, worked with us to create a clear and concise annotation guide, and consistently checked in to confirm they were still on track with each batch of outputs.

What do you dislike?

The process of submitting annotation guidelines occurred through a template form, which meant we had to recreate our existing instructions in the new format before uploading. However, this was only a mild inconvenience since they dedicated a team member to assist with the transition.

What problems are you solving with the product? What benefits have you realized?

Our university lab researches how to detect modifications of printed circuit boards by analyzing optical images of their surface. The process involves locating each surface-mount device (SMD) in the image (e.g., resistors, capacitors, ICs, etc.) and ensuring their specifications are within required tolerances. However, meeting the goal of component-by-component assurance involves training complex neural networks on a significant amount of labeled ground truth data. We reached out to several annotation services, and SuperAnnotate was the only one confident to handle large, high-resolution images with hundreds or thousands of complex SMD footprints. Through their platform and QA pipeline, we have obtained enough annotations to kickstart our machine learning processes and move toward the goal of hardware-level PCB assurance.