-
Notifications
You must be signed in to change notification settings - Fork 52
Contributions
What does this compare to? What is the difference between that? Are we just reducing the number of steps? Or getting you to understand how to adapt and modify an example?
Sharing and adopting seems most novel. How do we test this? This is somewhat orthogonal to the domain of Arduino / sensors.
How does sharing and adopting examples w/ this tool differ from how you'd do it otherwise? e.g. comparing neutralizing other hurdles (like uploading to Arduino, managing data) and just test the tweaking / tuning of examples.
comparison to just doing everything on the laptop. or comparison to National Instruments boards.
Question: what other sensor types? what can we detect?
Question: do we first need to re-build Exemplar or MAGIC to do this?
Question: is there a specific domain that we can start with?
Question: are we assuming a separation between example writers and users? vs. crowd-sourcing (where many people gather training data for someone's recognizer). regardless, there could be a way to share additional training data.
Question: how do you document the environment that collected the data?
One take on this is peer-to-peer repository of shared models and data.
Another take is an interface for tweaking and calibrating and modifying examples for your own needs.
Question: what can you actually do on a microcontroller?
Question: why use a microcontroller at all?
Part of the research is figuring out what's possible to do on a microcontroller.
Shielding from unnecessarily complexity, explaining the steps required.
Are we trying to teach people? Or just get their projects working?
Helping people think through how to get something to work.
Have to build those things into the workflow of the interface.
What's the insight we add by making this accessible to new audiences? Define pain points. Show how those are resolved by our system / examples.