Konvolutional Nöral Ağ Kullanarak Hasta Elma Ağağı Yapraklarinin Segmentasyon
Loading...
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Open Access Color
OpenAIRE Downloads
OpenAIRE Views
Abstract
Tarım alanında, uzmanın gözü hastalığı erken bir aşamada tanımlayamayabilir veya doğru bir şekilde teşhis edemeyebilir. Bitki hastalığının yanlış teşhisi genellikle yanlış tedavinin seçilmesine ve bu da mahsulün kaybına neden olur. Bu nedenle, hastalıklı yaprağın otomatik segmentasyon sistemi bu sorunu çözmek için son derece gereklidir. Bu tez Bitki Patolojisi 2020 segmentasyonunda derin öğrenme nin cesaretini görüntüler - FGVC7 veri seti elma kabuğu gibi birden fazla elma foliar hastalığı belirtileri yüksek çözünürlüklü renkli görüntüler içeren, sedir elma pas, ve sağlıklı yapraklar. Önerilen segmentasyon algoritması, U-Net ve ResNet olmak üzere iki farklı mimari kullanılarak yapılan anlamsal segmentasyon yaklaşımıdır. Her iki ağın sonuçları Pixel Accuracy, IoU, F1-Score ve Recall ölçümleri kullanılarak değerlendirilmiş ve karşılaştırma ResNet'in bu amaca yönelik verimliliğini göstermiştir.
In the field of agriculture, the expert's eye might not be able to identify or correctly diagnose the disease at an early stage. The misdiagnosis of plant disease often leads to choosing the wrong treatment and this leads to losing the crops. Therefore, an automatic segmentation system of the diseased leaf is highly required to solve this issue. This thesis displays the prowess of deep learning in the segmentation of Plant Pathology 2020 - FGVC7 dataset that includes high-resolution coloured images of multiple apple foliar disease symptoms such as apple scab, cedar apple rust, and healthy leaves. The proposed segmentation algorithm is the semantic segmentation approach by using two different architectures U-Net and ResNet. The results of both networks have been evaluated by using Pixel Accuracy, IoU, F1-Score, and Recall metrics, and the comparison showed the efficiency of ResNet for this purpose.
In the field of agriculture, the expert's eye might not be able to identify or correctly diagnose the disease at an early stage. The misdiagnosis of plant disease often leads to choosing the wrong treatment and this leads to losing the crops. Therefore, an automatic segmentation system of the diseased leaf is highly required to solve this issue. This thesis displays the prowess of deep learning in the segmentation of Plant Pathology 2020 - FGVC7 dataset that includes high-resolution coloured images of multiple apple foliar disease symptoms such as apple scab, cedar apple rust, and healthy leaves. The proposed segmentation algorithm is the semantic segmentation approach by using two different architectures U-Net and ResNet. The results of both networks have been evaluated by using Pixel Accuracy, IoU, F1-Score, and Recall metrics, and the comparison showed the efficiency of ResNet for this purpose.
Description
Keywords
Elektrik ve Elektronik Mühendisliği, Electrical and Electronics Engineering
Turkish CoHE Thesis Center URL
Fields of Science
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
0
End Page
77