full_width_weeds Computer Vision Project

We will try to retain the wide and short size of these images. This is how the Oak-D camera will send video to the NN model, wide and short, so that the weed spraying wand will be directed left or right only, for it dosn't have fore and aft movement yet. The machine moves forward, of course, and we plan that the speed will allow for good timing and spraying of the weeds. For this reason, only the strongest detection will be sent to Raspi and servo for moving the wand; there won't be time to spray two spots in an image. We will train in Yolo5 on Ultralytics, save to Openvino format, then convert that into the "blob" that Oak requires. The Raspi will also run the Python main script and control the relay and pump for herbicide application.

Cite this Project

If you use this dataset in a research paper, please cite it using the following BibTeX:

@misc{ full_width_weeds_dataset,
    title = { full_width_weeds Dataset },
    type = { Open Source Dataset },
    author = { Russ Hall },
    howpublished = { \url{ https://universe.roboflow.com/russ-hall/full_width_weeds } },
    url = { https://universe.roboflow.com/russ-hall/full_width_weeds },
    journal = { Roboflow Universe },
    publisher = { Roboflow },
    year = { 2022 },
    month = { oct },
    note = { visited on 2022-12-04 },
}

Source

Russ Hall

Last Updated

2 months ago

Project Type

Object Detection

Subject

weed-location

License

CC BY 4.0