Hello BSTRobin, hello everyone, hello Bosch – I followed the measurement of the types of coffee with the Bosch BME 688 Shuttleboard and trained the AI (4 specifications and 1 hour per specification). The detection of the samples also works in principle. But if I test the samples the next day with the same algorithm, then there are problems recognizing the samples! How come?! Do I need constant conditions? For example, does the weather affect my measurements?! And if so, how can I compensate?! Do I need more data to learn a specification, i.e. longer data collection times - and if so, can I do this with the AI on the PC and the mobile app, or do I need the AI server for this?? In short: what can I do to make my gas measurements independent of environmental influences (sun, temperature, pressure).Or is it precisely the concept of patterns in environmental influences, i.e. compositions of gases, that constitutes the design of the AI? I'm experimenting with the Bosch BME688 shuttleboard and would really appreciate any help! 🙄
... View more