A Concise Review on Deep Learning Methods in Agricultural Applications

Agriculture is one of the industries that plays a vital role in the development of a country. Industry 4.0 has led to the development of intelligent machines equipped with various sensors for automated monitoring. While sensor-based data collection is a growing field in agriculture, this review primarily focuses on the application of deep learning models like VGG16 and YOLO for image analysis tasks in agricultural sectors. Nevertheless, these large sensor networks produce a large number of data, whereby they need to be processed intelligently. A deep learning approach is currently the state-of-the-art technology that is able to infuse a large number of inputs to deliver high-performance intelligent applications. Deep learning models come in a variety of forms, including deep neural networks, recurrent neural networks, neural convolutional networks, and many more. This paper will focus on a review of the deep learning approach in agricultural applications through three specific key tasks: classification, detection, and segmentation. This review highlights the usage of VGG16 and AlexNet as a commonly employed pre-trained model for the classification task, the YOLO model for the detection task, and an integrated system of deep learning networks for the best segmentation task across several agricultural applications. One of the popular methods used to improve the base model performance is by integrating advanced module as well as pruning the whole model. In conclusion, this paper has contributed to a comprehensive understanding of the reviewed deep neural networks for agricultural applications. For future works, more works need to be done to continuously evolve the deep learning algorithms to help in various domains of agricultural applications.