Weed detection algorithm

Hi guys,

I saw the FarmBot Genesis introductory video on youtube, and there was one scene where the narrator talks about an advanced computer vision software to detect weeds. It really got my attention, but I couldn’t find anything about it. Could someone please provide me a link to the source code of this algorithm? Or maybe just sketch up a basic overview for me?

1 Like

Agree with you, I check every depot on git, I don’t understand how weed detection can work. Nothing on it.
I think it could run on Pi with some openCV algorithm or ITK. However as farmbot know where is the seed, everything “green” which isn’t near the seed could be assimilated as weed.

Hope we will have answer soon.


Yes, differentiating based on location will work most of the time, but what if a weed grows so close to the plant that viewing from above they are overlapping / give a continuos outline? I admit that this is more like an advanced weed detection as removing them will also be more challenging without hurting the plant. I think that this is an interesting problem and I am very interested in a solution for it, even if we have to revers engineer it :slight_smile:

1 Like

The weed detection code is on GitHub: FarmBot Labs Plant Detection.

@pcdeni Yes, our current method is to only mark plants as weeds if they are outside of the expected region of the desired plant. The few weeds close to the plant that escape detection can be carefully removed by hand to avoid damaging the plant.

1 Like

Welcome! Get rid of weeds rather dark film