QuiltView: a Crowd-Sourced Video Response System

The paper discusses a crowd-sourced system in an effortless one-touch video capturing using Google Glass hardware. The user receives queries visible on the Glass viewfinder relevant to their current location and opt-in (subscribed) preferences. This system allows micro-transactions (quick interaction sessions between the Glass and the user) that encourage the user not to get distracted from their situated context.

Main Points:

•           QuiltView is a crowd-sourced system that leverages the ability of wearable computers such as Google Glass near-effortless capture of a video in first person or point of view (POV)
•       How could QuiltView make a difference in real world?
•           Traffic Emergency: During a catastrophic traffic accident, users can record the video using Glass relevant to traffic congestion and incident to aid other user’s to re-route in order to avoid traffic congestion and help traffic authorities study the incident even before they reach
•       Missing Person: If a person is being abducted and nearby locals are not able to handle the situation, recorded video from the Glass can be made publicly available to help in further investigation. Crowd-sourcing techniques offered by QuiltView can help find the suspect faster
•       Real-time Restaurant Information: The users can upload video of the queue line and buffet menu for other potential customers who are deciding upon which restaurants they may go to
•       Time Machine: similar to the ‘missing person’ incident, the users can record and report a crime using Glass that helps law-enforcements to re-create the crime scene
•       QuiltView uploads the video to YouTube, recorded by responding users and relies on cloud to maintain a catalog of video meta-data that includes YouTube URL, location of video, owner, time, query and some other fields. Location are results of query from Google Maps
•       The system implements caching to avoid the cloud being flooded by multiple requests for the same location during events of high public interest
•       Once a query is submitted, for instance, a picture and details of a missing bike, the query includes location coordinates and zoom level using Google Maps in order to achieve an effective caching of results. This helps the system to recognize semantically close queries and enable caching of such queries. The paper states that categorizing related queries is an open-ended problem that may require natural language processing and other artificial intelligence and machine learning techniques
•       Load balancing in QuiltView’s perspective refers to the cognitive burden placed on users who receive queries. That is, the user gets to opt-in to specific queries based on the time and location and set parameters such as the number of queries the user is willing to receive on a daily basis. QuiltView identifies Glass clients based on these parameters and randomly chooses desired number of users that will receive the query
•       The focus of QuiltView is to take advantage of low user distraction and low cognitive load for interactions employing crowd-sourcing techniques.

Drawbacks:

•         Users who are not willing to take part in crowdsourcing or those who do not own a Glass may be recorded and uploaded by the system during an event or incident. This raises concerns for user privacy
•       QuiltView may be unable to take advantage of cached results and unnecessarily contact users. Taking away low user distraction feature offered by the system
•       The paper states ‘.mp4’ as their desired video codec. This type of codec is heavy to process and large in size. The glass will experience a great hit in battery performance during uploading. The paper should explore more light weight video codec options such as ‘.flv’
•       Streaming of YouTube videos require a constant stable connection with the Internet. This negatively affects battery performance on the Glass
•       Since videos are shared using YouTube API’s, making content available in real-time is cumbersome because YouTube doesn’t support such real-time services
•       Other than the reward feature, Glass users do not have any other incentive to reply to queries needing attention
•        The paper doesn’t discuss on how the Glass communicates with the cloud. Since Glass is only limited to using WiFi, it will need phone’s data services to receive and respond to queries. Cost will be a major factor as streaming and uploading videos on data services is quite expensive. In such cases, the reward may longer be an incentive for the user to respond to queries