Abstract:
Bangladesh is a six-season country and is also called a riverine country. New things are seen every six seasons, such as different types of flowers, fruits, pithapuli, festivals, seasonal animals, insects, etc. Seasonal changes such as river water level increases, demand for seasonal fruits is high, seasonal animals and birds can be observed, and the effects of various seasonal pests or insects can also be observed. As a result, many crops are destroyed by pests or insects, causing food shortages throughout the world. So, our work mainly aims to identify different types of insects or pests through artificial intelligence and data mining algorithms. The main objective of this work is to focus on agricultural practices so that farmers can detect and save their crops early as well as seeing how farmers can use less pesticides on their crops. As a result, the use of pesticides on the cropland can be avoided to a large extent, and the fertility of the land will remain the same. This means that farmers can protect their crops from harmful insects, they can use fewer pesticides on crops, and biological monitoring workers can save time. In this paper, three transfer learning models, namely VGG19, MobileNetV2, and ResNet152v2 and one deep learning model, namely Customize CNN, are used. VGG19, MobileNetV2, ResNet152v2 transfer learning models have 94% accuracy in VGG19, 90% accuracy in MobileNetV2, 96% accuracy in ResNet152v2. Also, the deep learning model Customize CNN has 80% accuracy. It can be seen that the ResNet152v2 model obtained the highest accuracy of 96% using transfer learning, and the Customize CNN model obtained 80% accuracy using deep learning. Here, it can be seen that the highest accuracy was obtained using the ResNet152v2 transfer learning model.
