@vroyer I agree that the weed detection functionality - like all parts of FarmBot - can be improved. Validation and filtering capabilities would be great, photo stitching would be awesome, automatic camera calibration would make things a lot easier, AI-based analysis has the potential to be a lot more accurate, cloud-based computing would have its own advantages (and disadvantages), and the list could go on.
Each of these things are non-trivial to develop, and will take most of our team’s attention for several weeks to several months to build. While we do plan to improve weed detection, those goals are parallel with many other goals we have, and as mentioned in other posts, not the immediate priority.
This is a little off-topic, but it seems you are very unhappy with where FarmBot is today - which is totally OK and understandable. What you seem to want is what FarmBot might be after 3, 5, or 10 more years of development. If you would like to return your kit to us, please DM me. We’ll offer you a full refund and pay for the shipping. Perhaps you can get involved again in a few years when the technology has matured and is closer aligned with or even exceeds your expectations.