The laboratory of Muse.
Not every question requires a product decision. Experiments provide a way to test small ideas, technologies, assumptions and approaches — and to keep both the successful and unsuccessful ones on the record.
Clothing recognition
ExploredQuestion: how accurately can a garment be recognized and described from a photograph taken by an ordinary person, in ordinary conditions?
Related: Google Cloud Vision exploration.
Visual product search
ExploredQuestion: can product-search infrastructure connect a photographed garment to useful product information — and is exact identification even necessary?
Related: Google Cloud Product Search, product-set approach for clothing images.
Wardrobe → outfit matching
ProposedQuestion: once garments are represented together, what actually makes two pieces compatible, and can that be computed?
Not yet started.
These are investigations in progress, recorded here so the questions stay visible even before there are answers. Results will be written up in full using the format below.
Every experiment is recorded the same way.
- Question
What are we trying to understand?
- Hypothesis
What do we expect to happen?
- Method
How will it be tested?
- Result
What happened.
- Conclusion
What the result tells us.
- Decision
What changes in Muse because of it.
- Status
Active, completed, inconclusive or abandoned.
Completed experiments stay accessible.
A failed experiment is still part of the project's history. The archive preserves both successful and unsuccessful approaches, because knowing what did not work is often the more useful record.
The exploration so far has been technical investigation rather than structured experiments with recorded results. As the investigations above are written up against the format, they will appear here.