Connected Devices Weekly Update/2016-06-28

• MVP
  • A device that creates shopping lists with voice commands
  • Integrate with a 3rd party list app (Evernote)
  • Can be shared with multiple accounts
• Upcoming Defined Milestones
  • Legal review - Bug 1280765 User testing form to sign complete
  • Nascent Objects Shipping Timeline Confirmed - Done (Personal communication with Baback Elmeih)
  • Confirm when devices will be in the SF office - Done (July 15-22)
  • Beta version of software (http://bit.ly/28YnCyK)- ETA July 1
  • Vaani internal testing - July 1 - July 15
• Milestones in Progress
• Milestones that need Defining
  • User testing that will happen with the devices - How are we going to be testing them and frequency
  • Users that are receiving a device
• Updated Lean Canvas

(new wiki is coming soon)

Introduction to Project Cue

• Based on research study results, the Link team pivoted from a hub product to a family organizer (bulletin board type functionality for schedule, reminders, lists) and is also exploring a smart camera. Team members have been identified for each of these two efforts and have begun work.
• First focus for the family organizer will be on reminders, with this effort called Project Cue. Currently, busy families are using post-its, their memories, notes, etc. to remember the who, when, what, where, and if recurring for things they have to do. The team's prototype will try to answer questions like the following:
  • Can we use voice to create these reminders?
  • Can we use them whether family members are home or away?
  • Can we provide reminder notifications for any family member?

MVP for Project Cue

• Voice
• Add reminders (for myself and others)
• Notifications (sent in 2 ways: voice TTS as well as app notification on device)
• Single account per family (everyone gets the reminders)

Prototype

• Team will take four weeks to develop a simple web app. that will test the following:
• Should have something working end to end for reminders by this Friday, if all goes well

(new wiki is coming soon)

Introduction to Project Sentry

• The focus of this smart camera project, called Project Sentry, is a low cost smart camera. People are looking for a way to monitor their home (kids at home with babysitter, baby sleeping, waiting for a package delivery). Current devices incur a one time high cost and then there are storage fees, so local storage and low cost are the differentiators.
• With this project, the team needs to find out if it can we build a low cost camera with low storage fees to break through these barriers.
• The team wants to build a camera that can:
  • Trigger snapshots based on motion triggers, stream video
  • Learn usage patterns
  • Learn storage patterns

MVP for Project Sentry

• Snapshot
• Livestream
• Remote access
• Notifications
• Purging rules
• Trigger camera action (motion as trigger)
• Storage (on device)

Creating an ambient method of communicating for 12 to 15 year old girls will create fewer interruptions that distract them from being in the moment at home and at school and improve their personal connections.

• team on PTO last week
• During this week we will continue working on getting the insights from the second User Research (that will help us to make the next decisions)
• Documentation work for building the prototype so anyone could build their own and adapt it to their needs. High-level summary in the README, and more instructions in a document - detailing each step in the device building, software flashing and configuration process, together with a 3-part series of step-by-step build videos

No new hypothesis yet

• Once our first hypothesis was not validated we decided to run some experiments in London All Hands to get some good insights that let us choose a single direction to follow. Data we are analyzing from the ww
  • Metrics: Dashboards created in redash with information about installations, launches, URLs expanded and swiped.
  • Metadata server: Requests to the server during the experiments
  • Shortened URLs clicks (bitly): What Tiles/URLs are more often clicked and impressed in Magnet
  • Feedback from the surveys filled that were broadcasted in the beacons
  • Interviews run from the people that left their mail address to be contacted
    • Probably we will send a questionnaire to the CD staff to have more input.
• Analysis is still ongoing but we have already detected:
  • Concerns about lack of any type of notifications from Magnet app
  • Setting Beacons is hard
  • Interest for broadcasting personal info and how to create content in an easy way
• We have already started to analyze the different types of notifications that we could use in iOS and Android (more info github and discourse)

• General updates
  • Having consensus on working process for Q3 (clearly list the experiments down with hypothesis, activities, metrics and host planning session to sort
• What’s done
  • MVP for Q2 approved by PIB
• What’s next
  • Wrap up Bug 1271206 (SensorWeb Q2 goals) and release sprint report by end of this week
  • Work on details of P1 experiments

• Trello: https://trello.com/b/b402OoGh/smart-kitchen
• Currently focused on getting a camera working with Raspberry Pi that can do both visual item recognition of food and barcode recognition
• As of late last week we abandoned the Raspberry Pi Camera (raspicam) and have moved on to a USB camera with Autofocus. We believe the autofocus is key for this project.
• Barcode recognition
  • Last week we successfully got the barcode reader working with the Raspberry Pi.
  • Utilizing SimpleCV library for the camera software.
  • Utilizing a free UPC database to recognize UPC-E barcodes
• Image recognition
  • We are now using the autofocus camera with SimpleCV library
  • Last week we created some custom classifiers, but the default ones seem to work better. This seemed interesting as the images were more easily recognized when they were lower resolution vs. higher.
  • This week we have had some issues with the service as my free trial has run out. The service is back up and running as of this morning.
  • As of this morning, I'm able to create a custom classifier for a green pepper, take a picture of it with the new autofocus camera on raspberry pi and submit it to BlueMix and have it recognize the object as a greenpepper with 52% confidence whilst the default classifier identifies it as a toilet with 33% confidence.
• Food Database identification
  • Evaluating taxonomical sources of food names. This will be key in isolating ingredients.
• Survey results - forwarded to research team. 70 responses on Busy Parent segment and 100 responses on Foodie/Millennial
• Design session planning for week of 7/11.

• Small changes in the library to make the Javascript API common for both client application (e.g. a website) and server side applications (e.g a node.js app)
• Helping Magnet team to be able to create his own dashboards from the data got from the experiments in the London ww
• The Mozilla Re:dash instance is now connected to our CD database so everybody can login with his @mozilla.com account directly
• Included more background info and a description of the CD tables/schema in the CD Metrics wiki

• Working toward a Q3 goal that will directly work toward supporting the top line Q3 goals of the CD org to provide user, market and technical research.
• Mapping out the whole team goals and objectives
• Defining what happens when people get pulled onto a project

• Working on post-London follow-ups with volunteers, 1-1 video chats, and surveys
• Creating an event toolkit for volunteers to run their own hackathons, innovation sprints, IoT meetups
• Formalizing criteria for community-run events
• Upcoming CD Events
  • Berlin, Germany - Mozilla IoT Tech Weekend
  • Amman, Jordan - Connected Devices Lab, University of Jordan
  • Hyderabad, India - IoT Hackday
  • Rio, Sao Paulo, Porto Alegre, Brazil - Vaani Hackathons