WATCH OUR PRODUCT DEMO HERE!!

Pitch deck link Google Drive link Video (YouTube) links Demo Part 1 link Demo Part 2 link Kalman Filter Demo link Heat Map Demo link

Inspiration

We tackled Problem Statement #5: Ghost Gear.

We identified with this problem as it is very applicable to Hong Kong waters. Hong Kong experiences an annual loss of $7.4 million through ghost fishing (2009 figures); the Hong Kong government has spent hundreds of millions of dollars of taxpayers’ money yearly, since 1998, to hire external agents to clean up marine waste on its shores.

Throughout the development phase, we tried to answer two questions: How do we identify ghost gear effectively? How do we trace ghost gear back to its origin?

The solution we drew up, Phishing for Good (P4G), is an integrated software and hardware product that allows NGOs, volunteers, tourists and fishers to quickly and simply trace back a piece of ghost gear to its origin. Our smart IoT solution (smart gear tags) coupled with advanced technology (specifically image processing and machine learning) allows us to relatively accurately predict data, which we can then translate on a high-level onto a real-time heat map.

Volunteers have an incentive to use our app, as the information input procedure is simplified and made intuitive. Local governments have the chance to engage in environmental conservation of their waters through utilizing our app. Large- and medium-sized fisheries are given the opportunity to engage in corporate social responsibility (CSR) and to claim increased profit, due to the decrease in the scale of ghost fishing - an unfortunate byproduct, among others, of ghost gear.

Ultimately, our predictive technology allows stakeholders to identify what we call ghost gear “prone zones” across seas, improve awareness of potential ghost gear-related hazards and increase opportunities for timely gear retrieval.

Phishing for Good (P4G) can address the suggestions made by the United Nations Food and Agriculture (UNFAO) by firstly seeking out partnerships with fishing gear manufacturers to deploy smart tags, then by liaising with local governments to encourage logging of ghost gear, both found and lost gear, through employing a “no-blame” policy. This is because we believe that aggregation of data deserves priority over penalizing of offenders.

In this way, we believe that with the cooperation of relevant stakeholders, we have in our hands a valuable piece of technology that has the ability to maximize its potential in providing real impact to the fishing industry.

What it does

There are three main functions: Report Found, Report Lost and My Gear.

Under Report Found, perhaps the key functionality, the app user fills in four key entry fields, namely: location found; type of gear; sub-category of gear; and gear color. Once we are able to secure buy-in from gear manufacturers by way of deploying smart tags on fishing gear, we can utilize image processing using IBM Watson to enable one-step information input.

Report Lost functions in a very similar manner. The My Gear functionality allows fishers and other stakeholders to keep easy track of their gear by registering gear details, or by logging the details of the smart tag. The smart tag deployed on nets and other gear incorporate simple hardware: a low-cost accelerometer that complements our app. The tag, along with real-time data of ocean currents, allows us to map a probability model of where the gear was first lost and its likely trajectory. We can display this information on a heat map that serves as a visual representation of all the features of the app.

These features allow fishers and other stakeholders to report both lost and found gear so that ghost gear hopefully spends minimal time in the ocean and also has a chance to be cleared away by clean-up agents.

How I built it

We built an Android application that allows divers and NGOs to report ghost fishing gear that they find. In order to facilitate the tracking of the smart tags which we incorporate in our idea, we created our own algorithm based on the Extended Kalman Filter with a gain which is regressed over various samples of traced paths through a Grid Long Short Term Memory network. The algorithm makes use of an Inertial Measurement Unit’s data integrated in terms of acceleration, and also cached forecasted water current data to gain a fairly accurate approximation of the smart tag’s position over time.

The hardware prototype was made using an Arduino and MPU6050 IMU for the dead reckoning calculations. We also made a RESTful API for the backend using MongoDB and Deployd & hosted on Heroku.

Challenges I ran into

As our solutions integrates software with hardware, one of the main challenges was to come up with a solid vision of our technology. In order to build a reasonable model, we had to decide specifically the functionality of each component of our prototype, and figure out, how to make them feasible.

In particular, we had to decide how to relate the mathematical model for predicting the possibility of having a particular piece of ghost gear in some area (heat map) to the software solution (platform) for reporting found items by volunteers and divers. In addition to that, we had to work out the value proposition of our solution for the users, the fisheries, and the government. Overall, the ideation process was the major challenge for our team.

Accomplishments that I'm proud of

We are proud of our teamwork: both the tech and business parts were well analyzed, so we were able to differentiate ourselves from other teams in all aspects: from our tech slant approach to our pitching.

It took some time for us to structure all the information in an interesting, pitch-friendly manner and to create a solution based on a complicated algorithm. However, as we allocated adequate resources to every challenge we aspired to solve, we were able to excel in all of our tasks.

What I learned

One of our major learnings was realizing and understanding real challenges faced by the fishery industry, specifically in Hong Kong.

Below are a few points of note: Despite the number of awareness campaigns and programs, most people are still unaware of the long-term consequences of these issues.

Interestingly enough, losing fishing gear can cause just as much harm as capturing too many fish.

Ocean currents are strong enough to move cement blocks the size of a tank. Therefore, ghost gears are likely to end up in places that are far away from the point where it was lost – posing threat to rare aquatic species.

A huge database does not necessarily mean it gives you what you are looking for. At the same time, look at data for long enough and you’ll probably find what you need.

What's next for Phishing for Good (P4G)

Phishing for Good (P4G) currently uses open source data collected from Adrift.org (for probability values of ocean current drifts at particular lat/long values) and earth.nullschool.net (for real time values). The future, as we see it, is to work around linking these data sets to improve the accuracy of our data model, as well as using image processing and machine learning techniques to further simplify the identification process.

In terms of implementation of the product with fishing nets and getting fisheries/manufacturers involved, we are looking to appeal to the government as suggested by UNFAO (United Nations Food and Agriculture Organization) to bring about a regulation for all involved manufacturers to adopt the proposed technology.

Ultimately, we aim to achieve our goals in addressing the pain points of the given problem: that is, to identify and trace back the origin of the ghost gear and, in doing so, use this data to direct volunteers and diving groups to search for ghost gear. This would eventually allow us to identify fisheries that are more likely to lose their gears and probably reach out to them to address the fishing practices, quality of gear etc. to prevent future incidents.

We are thinking of extending the functionalities of P4G to include a secondary market for mildly damaged and/or unclaimed, retrieved gear.

Built With

+ 6 more
Share this project:
×

Updates