The benchmark detail page allows you to view analytical and visual comparison data against the file they benchmarked.

Top bar action buttons:

At the top of the page in the action bar, there are two available actions that you can take from this page.

Processing Detail:

The processing detail panel will open and expose the associated processing information against this file and its AI models. This panel helps you understand important processing information regarding the processing time and the status of your benchmarked AI models.

Data Details:

If you want to view the entire file you have benchmarked from the Data Center application, click the Data Details icon to open a new tab and view the file from that OS application.

Engine Filtering:

At the top of your page, you can select from the available AI models you have used in this benchmark. Using the checkbox in the dropdown you can hide or show the models you wish to evaluate. Hiding AI models will remove them from the charts and comparison tools on the page.

Performance by Model Chart:

Depending on the AI capability you have benchmarked the Performance by Model chart will allow you to evaluate the performance of the model across different capability-specific metrics such as accuracy, word error rate, speed, and cost.

Engine Output Comparison Tool:

An example of the Engine Output Comparison tool for Video data

Using the Engine Output Comparison Tool, you can easily gain insight into each model's performance using this visual tool. Depending on the capability you have benchmarked, this tool will allow you to compare the ground truth or baseline you have selected against each model and provide a visual interface to truly understand if the model's accuracy is correct.

  • Compare transcription engine output side by side with your ground truth.

  • View the exact location of located objects, logos, and faces in video and image data types.

AI Model Details:

This section displays each AI model that has been successfully benchmarked with the file and associated model information such as build ID information.

Did this answer your question?