Tom Tom Festival 2019

Extracting semantic information from point-wise observations at scale is a complex problem usually solved by using traditional brute force spatial analysis techniques. These techniques employ rigid rule-based approaches that, though somewhat accurate, fail to understand cases outside the given rule set and are computationally expensive. 
To address these shortcomings, CCRi has explores new ways to efficiently analyze tracks by encoding spatiotemporal components in a joint representation in the video domain. In this presentation, we will show the usefulness of convolutional feature extraction compared to traditional kinematic analysis practices, and demonstrate extraction of behavioral information for maritime vessels in large regions in an effort to detect illicit activities. An example of such activities is the monitoring of fishing transhipment behaviors, where smaller fishing vessels off-load their catch to large reefer vessels to maximize fishing yield in international waters, severely depleting natural fishing grounds. The presentation will include a collection of animated visualizations to illustrate the aforementioned solutions. 

You need this ticket from Eventbrite to sign up: Applied Machine Learning Conference.

Over 1.2 billion people around the world lack access to electricity. In developing nations, this problem is especially acute as it limits participation in modern economy and culture. Improving the electric grid, however, is often logistically challenging in these regions because there is rarely a complete and accurate map of the existing electric infrastructure. This map is crucial as there is no way to make informed decisions on how to spend resources to improve the electric grid without it.
Toward solving this problem, we built a pipeline to efficiently map the high-voltage (HV) grid at a country-wide scale. This pipeline relied on both machine learning (ML) and our Data Team -- a group of eight professional mappers. The ML component processed satellite imagery across an entire target country and returned geospatial locations likely to contain HV towers -- the tall metal structures that support HV lines running for hundreds or thousands of kilometers. Our Data Team then overlaid this information on top of satellite imagery and used it as a guide to help quicken their mapping of HV towers, lines, and substations. With this overlay, they could focus their attention on high priority areas and avoid the tedious task of reviewing entire countries worth of imagery by hand.

You need this ticket from Eventbrite to sign up: Applied Machine Learning Conference.

Utilizing R to analyze a large open dataset of historical public transit ridership data via a modern suite of tools. To improve the affordable housing situation in Charlottesville, new high density urban developments are needed. Where to build these new housing units isn’t always clear, but analyzing public transit data helps inform that decisions by showing how the community gets around. Forecasting future ridership growth by bus-route gives community planners additional data to inform their selection of optimal locations for future developments in the city.
By communicating results via mapping visualizations and focusing on relevant community issues, this talk is aimed at anyone interested in urban development in Charlottesville. This talk will show off packages for time-series analysis (fable, tsibble), simple features mapping (sf) and data wrangling (tidyverse), but since the talk is results focused no coding knowledge or experience is required. Technical details of preparing data for time-series analysis and mapping spatial data, will be made available as part of the talk for audience that is more technically inclined and could be expanded on to another more code-centric talk if there is audience interest.

You need this ticket from Eventbrite to sign up: Applied Machine Learning Conference.

As environmentally-driven policy has shifted to incentivize renewable energy, large scale solar farms are being installed globally faster than can be reliably tracked by interested stakeholders. Photovoltaic solar arrays, or solar farms, can be clearly distinguished in aerial and satellite imagery, which makes solar farm detection a great candidate for deep learning. We have trained a convolutional neural network using Sentinel-2 imagery to detect the presence and extent of solar farms. As Sentinel-2 satellites have a revisit rate of 5 days on average, we are able to provide information in near real time. We have built an interactive web application that allows users to visualize and chart the growth of large scale solar farms. We will discuss how we prepared our labeled training data, how we trained a convolutional neural network, and show our model results. 

You need this ticket from Eventbrite to sign up: Applied Machine Learning Conference.