Konferans bildirisi Açık Erişim
Güzel Mustafa; Turan Bülent; Şin Bahadır; Baştürk Alper
The knowledge of weed numbers is very helpful for many studies due to minimizing weed harm
on the crops as well as knowing the weed species and classes. In this study, we used a deep learning
architecture that was capable of detecting some weeds to count the weed numbers instead of classical
manual weed counting methods. The pre-trained deep learning weight belongs to YOLOv5 which
is used in this study, can detect 5 different phenological terms (cotyledon leaves period, 3-5 leaves
period, pre-flowering period, flowering period, and fruit and seed setting period) of some harmful weeds
(sherlock mustard-Sinapis arvensis L., creeping thistle-Cirsium arvense L. Scop, and forking larkspur-
Consolida regalis Gray) in wheat production and other crops with 98% highest accuracy. StrongSORT
with the OSNet tool is used as the multi-object tracker. The weeds successfully counted from any image
resources (image, video, webcam, etc.) while avoiding recounting the same object by computer vision. It
plays an important role in the studies aimed to understand weeds population spread, resistance gaining to
herbicides by weeds, the economical threshold of weeds, etc. It also provides these parameters cheaper
and faster than classical methods.
Dosya adı | Boyutu | |
---|---|---|
Mustafa bildiri.pdf
md5:e01e6d988c1698b85ea8b3acff6b294e |
1.4 MB | İndir |
Tüm sürümler | Bu sürüm | |
---|---|---|
Görüntülenme | 57 | 57 |
İndirme | 239 | 239 |
Veri hacmi | 337.5 MB | 337.5 MB |
Tekil görüntülenme | 52 | 52 |
Tekil indirme | 200 | 200 |