11-22-2022 03:34 PM
Hi, when I train a model using BME-AI Studio to distingjuish Coffee vs Normal Air, it seems the algorithm only uses one heater profile at a time to train the model. Is it possible to train a model with 4 heater profiles combined and get a single accuracy value?
11-22-2022 07:01 PM
You need to change your heating profile and multiple heating profile is possible for our BME688 dev kit.
Here is the step.
1. Run AI studio.
2. Click Configure BME Board
3. Select 4 heater profile, then click save to file. (It will generate bmeconfig type file.)
4. Replace original bmeconfig file in SD card with new one.
5. Collect data
Then, when you train your data you can see 4 results for each heater profile.
11-22-2022 11:23 PM
Yes, I know how to change the config file to select different heater profiles. But when I train the algorithm, it only uses ONE heater profile at a time. See screenshot. It shows that the accuracy for 301 is 77%, for 321 is 76%, for 322 is 94%, and for 323 is 82%.
But I want to train the algorithm using ALL FOUR heater profiles together. Is this possible? The hope is that by combining multiple heater profiles together, you can do better than using any single heater profile alone.
11-23-2022 10:57 PM
Currently, it's not available to train different profiles altogether.
For your case, 322 is the most accurate one, but I recommend to collect more data and reach to over 95% accuracy and less than 5% false positive.
Thank you 🙂