Nov 23, 2019 23:00:00

A company aiming to build a “poop image database” by recruiting poop images appeared

by

Alexas_Fotos

Researchers sometimes collect a large amount of data when studying a specific subject, but sometimes people often recruit things that people do not expect, such as “ photos of a penis ”. A health-related startup called Seed is creating a stool image database by collecting 100,000 “stool images” from people.

The world's first poop database needs your help-The Verge
https://www.theverge.com/2019/10/29/20937108/poop-database-ai-training-photo-upload-first-mit

In the United States, one out of five people suffers from chronic bowel disease such as irritable bowel syndrome, and Seed believes that building a stool image database may develop a system that can help these people. The Seed's research team will provide a tool that allows users to diagnose their health from stool images so that they can manage their health.

Seed co-founder and CEO Ara Katz said, “We don't always see stool as a source of information, but stool will certainly output intestinal health directly,” The Verge Talking to

To cooperate with the construction of Seed's stool image database, access Seed's ' Stool image transmission page ' on your smartphone.

Tap the “#GIVEaSHIT” button at the bottom of the page.

Then the submission form will open, so enter your email address and the time you normally stool in the form. Then check the box and tap “TAKE OR UPLOAD PHOTO”. If you have already taken a photo, upload it as is. If you have n’t, take a photo and take a photo. The photos taken will be used anonymously by Seed's research team, separating the email address and potential metadata that can be identified by the individual.

Seed's research team seems to have seven specialized gastroenterologists who analyze stool images sent from the general public, and gastroenterologists diagnose and classify all stool images That's it. Classification is done according to the criteria of

Bristol Stool Scale , from hard ones such as `` hard as nuts '' `` sausage type but rugged '' `` rough and irregular shape '' `` completely There are a total of 7 types, ranging from loose to liquid.

The stool images tagged by the doctor are then used to train the artificial intelligence model. They use a training system similar to that used by automated driving cars to identify trees and cats on the street, and allow artificial intelligence to check the status of stool as a doctor does.

The final purpose of this stool database creation project is to make a large amount of stool images that are usually difficult to obtain open source so that they can be used for academic purposes. David Hachuel, who collaborates with the database construction project, said, “People with chronic bowel disease are struggling to choose what they should eat and how much they should exercise. “It ’s important to develop a monitoring tool that can monitor the condition at home.”

by Pixabay