CNN analysis for Arbimon’s platform
Created CNN outputs for Arbimon’s platform
Role:
UX Designer and Researcher
Team:
Senior designer and me
Duration:
4 months (Jan 2024- Apr 2024)
Client:
Arbimon
Problem:
Scientists working with acoustic data from rainforests needed a user-friendly interface for Arbimon, an acoustic analysis SaaS platform, to efficiently run Machine Learning models, specifically Convolutional Neural Network (CNN), validate predictions, and export results without requiring extensive coding expertise. The output must streamline the process of filtering, confirming, and rejecting predictions for species' presence or absence, enabling scientists to focus on their research and conservation goals.
Solution:
Create a new page for CNN projects and outputs on the Arbimon SaaS platform. This new platform, based on our research, would have a list of projects that the CNN model can be run on based on the data, and make validations on the data.
IDEATION
We brainstormed a range of ways to exhibit the information for the output of the CNN algorithm on the data sets. The first iteration went through usability testing with various scientists who use the platform.
Challenges:
Coming up with the initial designs was very difficult since I had no idea what CNN was, what the output could look like and we did not have any prior models at the company.
Changes in business needs of the product.
Solutions:
Conducted a great deal of secondary research first to understand CNN and performed some competitive analysis. Secondly, I conducted some primary research in the form of interviews and spoke to the internal science team to provide me with sample data and output to get a deeper understanding.
I started off with the initial brainstorming by creating wireframes of a new CNN job and the initial summary page.
Then I worked on wireframing the filters and export functionalities and their various user states. In our standup meetings, I got feedback as seen on the sticky notes and notes around the wireframes.
Q1 Quick Fixes
Once I completed the rough sketches in the form of wireframes, I got to work in Figma to create high fidelity designs.
NOTE: We did not do mid-fidelity then high fidelity since the dev team needed the designs as soon as possible for deployment.
Learning: During this stage, I got to collaborate with engineers and QA for my designs. I was able to advocate for my designs decisions in this capacity and in some areas make changes to my designs to suit the dev needs. This was a really important learning moment for me as I was able to build confidence in my skills and my designs for the project.
Below are some of the Q1 designs.
This page shows the options that a user has to create a new job to run the CNN algorithm
This page shows the summary of a job that is in progress and all the detections and statuses of the project.
This has the calendar which was a drop down in the wireframes but this is easier to use and visualize.
This page shows the recordings for each detection by dates and times. The validation status is supposed to filter the recordings.
Usability testing and Shipped Designs
The final designs came as a result of usability testing with scientists who would use the CNN algorithm. I got feedback on the filtering:
Q1 designs for the filter functionality was crowded and instead would be better to have a pop up perhaps for ease of use.
Scientists also wanted the option to use saved filters since their daily work is around similar filters.
The users wanted to see examples of the different export options to get an idea of which one to choose.
Page showing the ‘Export’ functionality in the summary of a CNN Job Page
Page showing the ‘Filter’ functionality based on the usability testing feedback and showcasing the feature to play a recording
KEY TAKEAWAYS
Advocate for design
At this internship, I was fortunate to ship products and hence I got to work the dev team, QA team and CTO. I was able to advocate for some of the design decisions I made based on competitive analysis or user research and this made me more confident in my skills.
Manage timelines and quick designs
Since the Q1 deadline was very quick, this was something new for me to work under a time crunch and do quick designing. This was really helpful in showing me the real world design work in a start up environment.