Metroparks Hackathon Information Session
Advanced Interfaces - Floor 2 at Sears think[box]
Richey Mixon Building 11201 Cedar Ave., Cleveland, OH 44106, United States
View Map
10
Registered
Registration
Registration is now closed (this event already took place).
Details
Track 1: Wild Plant ID App
This currently available iOS app was developed to help people familiarize themselves with the plant species near Euclid Creek. The project seeks to assist volunteers who are identifying plants as a part of conservation studies, and to engage more citizens in the natural environments around them.
- Provides a decision tree/matrix to help reduce the number of options when trying to identify species
- Provides photos of species in different contexts
Challenge
The Friends of Euclid Creek Watershed (the volunteer organization maintaining several of the conserved areas near Euclid Creek) will begin this year's field data collection in April. How can the app be more useful to the volunteers in these projects? How might the app address problems in citizen science in botany?
- What feature sets can be built in to the app to help users use it more often?
- How can the app be used to provide a similar experience for other geographic locations (different parks in the area, etc.)
Desired Outcomes
- Gain potential pool of developers to help on project
- Development plan for future updates (ideas for new uses for the project)
Track 2: Machine Learning
Cleveland Metroparks uses trap cameras through-out their twenty-three thousand acres of parklands to identify and track animal species and populations. The images, data and analytics generated from this ongoing project are used for several purposes:
- Collaboration with higher education institutions from a natural resource perspective
- As input into programs and projects that are funded by taxpayer and grants
- Identify shy or hard to see species (e.g. gray fox)
- As indicators of ecosystem health and thus an input into project development and funding needs in order to optimize resources for projects.
- Over the years the naturalist team that is leading this project has used a variety of methods to identify and classify the pictures taken. The approaches have ranged from crowd-sourced manual methods to the current methods using machine learning (ML) tools (Tensorflow).
Challenge
While machine learning has improved the identification and classification of pictures the procedures to make it work are still too cumbersome and manual. Additionally, the accuracy of the model outputs is not at the desired level to be useful. The challenges include:
- Large repository (11M+) of images and growing (3000+ daily)
- A wide variety in fields of view, lighting, and level of detail on video surveillance camera exports.
- ML algorithm accuracy to properly identifying images that should be excluded/removed.
Desired Outcomes
- Better model accuracy based on animal/no animal as well as species identification
- Faster data set processing
Track 3: 360 Video & Virtual Reality
Cleveland Metroparks has created a decently-sized library of 360-degree video content of locations and experiences at the park – these are currently passive experiences that users cannot interact with.
- Used with partner organizations to provide nature experiences to people who can’t get out into nature
- Used at Nature Centers for ADA accommodation
- Public 360 Video app forthcoming
Challenge
- We want to extend our VR experience to reveal unseen aspects of the natural world
- As much as possible we want to use natural/real assets to educate
- Animation is fine – we just don’t want a cartoon experience
- Use VR to play with scale (small as a bacteria or big as an ecosystem)
Desired Outcomes
- Educational VR experience that allows users to interact with and learn more about the environment they are in.
- Life of a Water Drop
- Woodland ID/Systems
Where
Advanced Interfaces - Floor 2 at Sears think[box]
Richey Mixon Building 11201 Cedar Ave., Cleveland, OH 44106, United States