This is a failed attempt to create a lightweight and opensource Hotword Detection model using LSTM.
The results that we can see were promising, as we can see in the images, the positive word(+) was detected and all the other ones, negatives(-) don't get a high result. But in practice, it was not accurate enough.
In the future I´ll probably try to make it work correctly.