Extracting information from Picasa & OneDrive for Cortana COMPUTER SCIENCE DEPARTMENT Technion - Israel Institute of Technology Extracting information from Picasa & OneDrive for Cortana Final Meeting Students: Ron Saad and Eliran Weiss Supervisors: Adi Miller and Michael Sterenberg 25.07.2019 CS 234313 Technion
the team Adi Miller and Michael Shterenberg Eliran Weiss and Ron Saad Microsoft Supervisors Eliran Weiss and Ron Saad Technion CS Students 25.07.2019 CS 234313 Technion
Goals Extract photo metadata from users’ Picasa and OneDrive accounts. Process the data and find trips the user has taken with various metadata features. Implement a tunable and configurable algorithm for experimentation and visualization. 25.07.2019 CS 234313 Technion
Methodology Our project was research-driven We started off with a vague concept of what we should accomplish. We’ve studied various public APIs and common user usage of cloud services. After we realized the capabilities of each API, we decided (with our advisors) that trips-inferring would be the best course of action. We’ve then studied several clustering algorithms to use in our own trips-inferring algorithm. 25.07.2019 CS 234313 Technion
Methodology (Cont.) After we’ve finished All research, we focused on coding We’ve build a small WPF tool to test our initial algorithm, with a relatively small amount of data. We’ve built a web application that saves the user’s data and shows him a textual representation of trips we discovered with semantic names (e.g. New York instead of 34.321 79.543) and time span. After we gathered enough data, we tweaked our algorithm and developed a web-based tool to tune the algorithm parameters and visualize the results on a map. 25.07.2019 CS 234313 Technion
Achievements A modifiable algorithm for trips inference. A web application that shows users their trips as a textual result. A web tool to tune the algorithm and view the results on a given user. We’ve gathered valuable data: We have a database of over 700,000 photos meta-data from over 100 users. 25.07.2019 CS 234313 Technion
Examples 25.07.2019 CS 234313 Technion
25.07.2019 CS 234313 Technion
conclusions The idea of recognizing trips with merely photo metadata is indeed feasible. We successfully differ between trips and short visits, casual hangouts, workplaces and home. More adapting and parameterizing can be done if the developer wishes to. 25.07.2019 CS 234313 Technion