User Testing Sessions
We held 2 user testing sessions internally with two of our team members acting as VPG engineers. Their feedback was used to iterate on the front-end prototypes of the models resulting in an effective high-fidelity interface.
Session 1 - Users were given tasks to complete using the low-fidelity prototype
Feedback was used to iterate on the low-fidelity prototype and transition to a high-fidelity
Session 2 - Users were given tasks to complete using the low-fidelity prototype
Feedback showed that the high-fidelity prototype of the front-end interface effectivelty met the goals defined during needfinding
User Testing Sessions #1 - Low-Fidelity Prototype
Users were given tasks to complete to see if they found the user flow intuitive.
Tasks:
Upload videos
Select a video
Choose a key frame for the model to analyze
Run the model
View model outputs
Key Insights
Flow was intuitive
Difficult to view key frame time stamps
Some action buttons were unnecessary
Users want to compare results between key frames and models
User Testing Sessions #2 - High-Fidelity Prototype
Users were given tasks to complete to see if the changes met their goals.
Tasks:
Upload videos
Select a video
Choose a key frame for the model to analyze
Run the model
View model outputs
These tasks were repeated for the Benchmark model, TSL Algorithm and CNN. Another set of tasks were adding to test the user flow of comparing the outputs of key frames/models.
Tasks for comparing key frames in one model:
Select a key frame
Run the model
View model outputs
Select a new key frame
Run the model
View model outputs
Tasks for comparing key frames between models:
Select a key frame
Run the model
View model outputs
Choose a key frame output that you want to use for comparison
Switch to another model
Select a key frame
Run the model
Compare model outputs
Final Feedback:
Users found the flow intuitive and agreed that they can clearly view model outputs and make comparisons.