Overview
Roboflow’s Universe is a dedicated to open-sourced projects where developers can contribute to computer vision projects, test and try pre-trained models, and share it with the community. It’s been growing rapidly and the number of contributions have increased to a point where navigating the universe has brought in some challenges. What are some improvements we can recommend to their design and structure to help with their growth? How can we improve the user experience and content organization that helps all users understand, create, and find computer vision projects more easily?
Goals
Design an interface that allows content to be easily categorized, discoverable, and scalable.
Objectives
Design a system that allows for a growing user base and the number of projects submitted
Make content organized and easily discoverable
Foster and grow a community around computer vision
Requirements & constraints
Use existing branding and colors
Make decisions based on available information and research
Only focus on Roboflow’s Universe platform
Research: data, stats, and insights
100K project submitted -> 500K+ projects submitted (growth in 9 months)
350M images
100K pre-trained models
5 project types
5 models
34 categories
Current UI of Roboflow's Universe community platform
Pain points and questions
I spent a couple days testing the current platform to understand the challenges of it's current interface.
Title’s do not give a good description of the intended use case or project
Images do not provide at-a-glance reference to what they project does
Are the meteadata and images relevant ie do users use these to search using these filters to find other similar projects?
Hard to determine what project is popular or useful beyond the star icon
How are projects grouped within the Universe?
User research and personas
I defined 3 main users of the Roboflow universe through research and insights based on looking through the Universe platform community, its users, and by studying their behaviors.
Likes to create projects in open source
Looking to buy/implement Roboflow for their company.
New to computer vision and Roboflow
Redesigning search architecture & content hierarchy
Through research and testing the usability of the current interface, this is what I think to be the current structure of the Universe (at a high-level).
The problem I found is that all 500K+ projects are independently free flowing within the Universe. They have tags, or identifiers that you can use to help find specific projects and some projects have been picked out via editorial content or grouped into favorited or top projects, but searching relevant projects within the current interface and structure is difficult and overwhelming for the amount of projects there are.
The new proposed structure will introduce some grouping and hierarchy to content at the highest level, making projects more organized. So instead of searching 500K independent projects, you can first structure your search to based on the category, then filter within that category by project type, model, etc.
This search system is beneficial because it works at the category level or within the entire universe itself, but gives the flexibility and ease for users when searching for projects.
Wireframing & Design
With the above information and research in mind, I designed wireframes that would address the following:
Hierarchy of content to make searching and discovering projects easier and more organized
Design the interface to foster a community of computer vision enthusiasts
Community driven interface
Since the Universe was designed to be contributions to open source projects from the community, I designed the platform to cater to those needs vs. just being a database.
Community first interface allows projects to be commented on, searchable, and highlighted in an organized and efficient way, and allows the platform to be a place for collaboration, engagement, and communication on all topics + projects related to computer vision and AI, which would benefit not only in managing the growth of the number of submitting projects, but also in growing a positive community of computer vision enthusiasts.
Components for educational purposes, featured projects, and category (ie community) organization
I designed 3 new UI components that will help with the user’s journey through the platform.
Left side bar (categories or communities section + general forums for Q&A, feedback, and FAQ for easy and quick access
Main feed /all (home feed for all projects, posts, submitted on the platform).
Right side bar (educational and quick links housed here - quick navigation to editorial content and documentation help to get someone started.
Learnings
Since this was only a look into their Universe platform, there are a lot of other factors I have not considered that could potentially change the design and interface entirely. There’s also some other technical considerations, resourcing, and business prioritization that would have to be understood that could alter or inform new design directions or solutions.
How is the workspace tool integrated into the user flow of submitting projects into the Universe?
Edge cases: What if a user doesn’t upload a gif or image for their project? What if they don’t input a detailed title for their project? (User testing and a further deep dive into the user journey and the structure of the current user flow will help me with these edge cases.)
How many times is an existing project is updated?
How are tags used and what's the frequency (ie is it something users tend to click to find relevant projects?
Exact user flow: Do user filter by number of images first, or by project type or model type first? Do they search using the metadata or objects of interests?
How often is an existing project viewed? How many DAU’s and MAU’s for the Universe platform, and what feature has the most engagement, interaction, or activity?