EyeSea

AI | Video Analysis | Human-in-the-loop

EyeSea uses machine vision algorithms to “watch” video footage for any incidents where a fish or mammal is near a MHK (Marine Hydro kinetic) turbine. It significantly reduces labor time.

The Needs

A common and simple means of seeing how fish and mammals interact with MHK systems is to set up an under water camera and begin recording. However, one hour of video can take 5 or more hours to analyze manually. This is a major bottleneck when hundreds of hours of footage need to be reviewed. Using machine learning algorithms to analyze the footage would greatly reduce the labor, but the accuracy could be unreliable. The app must also allow human annotators to override the results.

My Contributions

I was the only user experience designer in the team of researchers and software engineers. My contributions included user requirement gathering, user experience design, visual design, illustration and logo design. The application greatly reduces the average underwater video analysis time by 76%. 100% of the researchers within the organization use this application for their video analysis work.

The App Flow

Based on the user research, I carved out a happy path for the user:
Upload a video > select the right algorithms > inspect the machine detections > override the detections and/or add annotations > review and download the final result.

Clickable prototype

https://invis.io/75JGSVMM3ZG

The Iterations

Landing page

Before
The original landing page was not visually appealing and doesn't provide much information for the first time user.

Final design:
I designed the illustration to indicate the app's purpose and how it works.

Video overview layout

The initial layout: video overview on the right, detailed analysis on the left. The users recognized the purpose is to keep track of the progress of other videos while inspecting object detections; however, they preferred to have more screen real estate for annotating detections.

Final design:
Based on the feedback, I changed the video overview in thumbnail fashion. Click on a video to get to the detailed analysis view.

The detailed analysis view

Wireframe

Final design

The video detection summary

View Other Projects

Sensalert

Wearable | IoT | Health Tracking | First Responder

GridPIQ

Power Grid | Renewable Energy | Screening and Analysis Tool | Data Visualization

WatchOwl

AI | Data Visualization | Social Media Analytics