A computer algorithm to identify fish: A Q&A with ErieHack finalist Josiah Olson
Source: Peter Krouse
Josiah Olson developed a computer algorithm to differentiate types of fish.
His website is called whatismyfish.net, one of nine fledgling technologies competing for cash prizes at the ErieHack competition in Cleveland this week.
The idea sounds weird. But it makes sense for Olson, whose earliest memories growing up in rural Ohio are of fishing off a dock, catching bluegills on pole with a string and a big earthworm on the end.
Fishing stayed with Olson -- a Progressive Insurance programmer -- through high school in Licking County and college at Kenyon College, where he majored in mathematics and economics and developed an interest in data science.
Now that his team, Fish.io.ai., has made it through two preliminary rounds to reach the ErieHack finals, cleveland.com fished his brain.
Take a look.
How old are you?
25
Where do you live?
Euclid
How did you find out about ErieHack?
From an email blast, he said, "and I thought it sounded real cool."
He then learned that a representative of ErieHack would be speaking at the regular meeting of the Cleveland Area Python Interest Group at the LeanDog offices in the old Hornblower's restaurant barge on the Cleveland waterfront. He decided to go.
What is Python?
Python is programming language used by data scientists who want to retrieve information from large sets of data. Olson works with Python at Progressive Insurance to conduct what he calls "machine learning" or "predictive modeling."
Why did you want to apply predictive modeling to fish?
Olson said he fly fishes when he can, but that he finds it a challenge to comply with all the state regulations because he can't always identify what fish he has caught, and also because the regulations are not uniform and may differ depending on the fish, the time of year or the location.
He said he came up with the idea for a computer algorithm to make fishing regulations more accessible, but until ErieHack he didn't have the motivation to make it happen.
Why get your co-workers involved?
Because a lot of legwork is required. Whatismyfish.net uses the same type of machine learning that allows Google to organize images, he said, and to be able to accurately identify the 17 species of fish his group is featuring, they had to create a data set of 30,000 images.
Whatismyfish.net focuses on fish in Ohio and Michigan, but could be adapted to the whole country.
What fish are among the 17 species?
Smallmouth and largemouth bass, perch, carp, catfish and walleye are just some of the species that whatismyfish.net is designed to identify.
How does the technology work?
Using pattern recognition. Someone catches a fish and takes a picture of it. The picture is then downloaded to whatismyfish.net. The program learns how to recognize patterns from seeing lots of examples in the data set and therefore it can distinguish one fish from another.
Did you expect whatismyfish.net to have made such an impression?
"Honestly, no. I'm pretty shocked," Olson said. Initially, Olson viewed the algorithm as only useful to fishermen wanting to know more about regulations, but after talking to others he now realizes that it can be a valuable tool to organizations such as the Ohio Department of Natural Resources for tracking invasive species.
"I didn't really think how useful it would be to other people or the environmental impact," he said.
ODNR wants people to email the department if they catch a certain invasive fish, he said, but nobody does it. Maybe taking a picture and uploading it to whatismyfish.net will be more appealing.
How much time has your team devoted to whatismyfish.net?
Probably at least two team members have been working on the project at least four or five hours a night for at least the past three weeks.
How many computers are in your house?
"I actually only have one and its like really old and junky," he said, adding that you don't really need a fancy computer these days because you can essentially rent computing time in the cloud for pennies an hour.
How often do you fish?
Olson tries to get out at least once a month in the spring and fall when the fishing is good. "When your job is working indoors it's really great to get outside and just think for awhile," he said.
What are your favorite fishing spots?
The Chagrin River and the Grand River, but also the many streams in the vicinity of Geneva.
Who are the other members of the team?
Becca Rice, Matt Graziano and Matt Sochor. All work for Progressive Insurance.
| }
|