A demo application of Plant disease detection model built with Streamlit library. Here's the demo
I used this dataset from kaggle for plant images. For non plant image detection, I used a private dataset for non plant images. See the images below for a detail description of dataset and augmentation techniques used.
To solve this task, I have created a model that includes detection of non plant image as well as detection of excluded plants. Currently the model can detect citrus, tomato and potato plant diseases and ignores other plants. In total, the model has 19 classes. The non plant images are added as "not_plant" and excluded plant categories as "Other_Plants". The rest are healthy and different diseased plant images. For details of the class labels, see the static.py file. \n
I fine-tuned Efficientnet V2-B2 architecture on the custom dataset. See the model metrics below
