I did everything in accordance with the manual "Getting started"(Example: Coffee vs. Normal Air). I used all the default settings.
I have successfully exported the BSEC Configuration File and included it in my application.
My coffee is recognized correctly(BSEC_OUTPUT_GAS_ESTIMATE_1=0.99).
But many other smells (e.g. tea, perfume, flowers, etc.) are also identified as this coffee(BSEC_OUTPUT_GAS_ESTIMATE_1=0.98..0.99).
BME AI Studio: 1.6.0
What am I doing wrong?
Solved! Go to Solution.
If the gas curve of other smells are similar to coffee, it is likely that other smells can be identified as coffee. You could try to label more different smells and train the data to see if different smells can be classified.
Thanks for the answer.
The gas curves are significantly different.
We trained the algorithm to recognize air and coffee. We now provide tea to the algorithm.
The tea curve differs significantly from the air and coffee curves.
At the output we received:
Thus, tea was identified as coffee.
Is this normal behavior?
If I understand correctly, any "unknown" smell will be classified as the most similar class(it will probably be coffee). But why 0.99?
Can we somehow filter out "unknown" smells?