Add single class

Part 1 (2019)
1

Hi
rookie question:
How you retrain a model to add detection of a single class
(company tee shirt)

do you need examples of “NOT” the class,
wouldn’t that be everything else in the world ?
i since found this https://medium.com/@lankinen/one-class-classifying-what-kind-of-data-set-i-should-have-1486358e491b
which was great but still interested in anyone’s thoughts.

Thankyou

Thanks

2

Are you trying to classify images that always have shirts as company tee or not company tee? If so, you need images of the shirts you’re trying to find vs other random images of shirts you’re not trying to find. Otherwise, you’ll need a few other random photos to prevent your model from guessing “company shirt” 100% of the time. It’s not going to recognize other things from the random images, it’ll just learn to ignore things that it can’t rely on when it comes to detecting company shirts. Hope I made it clear

3

Hi Mike great question!
I hoped to detect the tshirt in a group of people or just one person.
The alternative would be similar groups wearing general clothes, including other tshirts.

Does that make sense?
Perhaps the scope is too great.

Thanks for answering.

4

You’ll be able to pull it off. I think object detection Will work better than image classification for that though.

5

– Yikes I see object detection is way ahead in course ; )

6

Image classification will certainly work, too. It just might get confused if there’s noise in the picture (multiple shirts, for example)