December 20, 2017 With a Crowdsourced Algorithm, NMIO Can Now Fight Illegal Fishing

The Topcoder Community has been solving data science problems for years — ranging from developing apps focused on social good to developing smarter cancer recognition technology. This fall, the National Maritime Intelligence-Integration Office (NMIO) completed an incredible project with Topcoder. We ran a crowdsourced algorithm competition in two parts to develop a method to enable authorities to identify and combat illegal, unregulated, and unreported (IUU) fishing activity globally. More specifically, the algorithm needed to be able to successfully fish for fishermen — to detect whether a boat was fishing or not fishing.

Fishing for Fishermen: the hunt for illegal fishing activity

During each phase of the competition, we ran a marathon match, a type of challenge we use at Topcoder to work on difficult problems in AI, algorithm optimization, and image processing. In a marathon match, contestants write their code, compute their data, and submit it for scoring over a period of 2-3 weeks on average. Competitors’ results are then posted to the public leaderboard for the rest of the Topcoder Community to see. For the Fishing for Fishermen challenge, we ran two marathon matches with 133 total competitors, 14 of whom competed in both marathon matches.

Marathon matches I + II

The goal of the first marathon match in the first phase of the competition was to identify which vessels in 10 pre-selected areas were fishing. For this round, members used satellite-collected vessel position information pulled from Automatic Identification System (AIS) messages. Seventy-seven competitors participated, and Psyho, a Topcoder member and copilot from Poland, won $5,000 for his winning submission.
The second marathon match had a more complex goal: to determine what types of activities vessels were engaged in. For this round, exactEarth AIS data was provided by the Harris Corporation, in addition to the data previously provided by ORBCOMM; our partner DigitalGlobe also compiled and provided oceanographic and bathymetric data from original sources for this round. Fifty-six competitors submitted solutions for a prize of $10,000; the winning submission came from wleite, who works in forensics for the Brazilian police by day and solves data science problems on Topcoder by night.

Actionable results from the Topcoder Community

Ultimately, the Fishing for Fishermen challenge — co-sponsored by the U.S. Departments of Defense and Homeland Security, along with the support of Harvard’s Crowd Innovation Lab — brought 119 Topcoder Community data scientists from 33 countries together to submit their unique solutions and ultimately help conservation and reduce environmental harm. Through cost-effective crowdsourcing competitions, the Topcoder Community was able to develop a process that can effectively identify and combat the global threat of illegal fishing.
Ready to solve your data science problem with Topcoder?