Object Detection models training

[koldogut]

I'm using the default object models in deepstack and sometimes I get false positives detecting persons (latests with a cute snowman in my garden at 04:55am, hehehe); as my home alarm is in part based on this detection i'd like to tell the model when the coincidence is true or false. Any suggestion ?

[robmarkcole]

I dont understand the question, you want a model that rejects snowmen?

[koldogut]

Hehe, no, I want the model to not identify the snowman as a person. I understand it should be done by re-training the model to get more accuracy but I don't know how.

[robmarkcole]

Please raise this for discussion on the deepstack forum, see you there