A lightweight deep neural network implemented on MATLAB without using GPU for the automatic monitoring of the plants

Plants and trees are essential for the human kind which must be preserved and well taken care for the growth of any country. Plants require intensive care which involves monitoring of their size, growth, health, and yield. Manually monitoring of these factors is time consuming. In this regard, computer vision approaches are evolving to monitor the plants for developing better yields. Deep learning convolution neural network is the most promising and currently trending tool which can solve the complex problems in the field of image processing and computer vision. However, often these deep learning models are computationally expensive and time-consuming which makes them unsuitable to be used in the real world scenarios. Thus, a lightweight deep learning convolutional neural network was designed and proposed to automatically segment the plant’s leaves so that the monitoring of the plants can be done with high speed. Benchmark datasets in four different categories with total of 810 images were collected to train and test the proposed model. Many challenges were associated with those categories under the mentioned dataset for the segmentation of leaves. The proposed light-weight deep learning model was implemented on MATLAB platform tool and trained on a local system without using any GPU (Graphical Processing Unit). The training was completed within 13 minutes and 25 sec and provided segmentation accuracy on testing set as 98.26% which is satisfactory for such kind of monitoring applications.