Implementation of Plant Seedling Classification Framework Using Residual Inception Network

One of the important and challenging fields of research in plant science is monitoring seedling growth and its early development. Characteristics including texture, colour and shape have been used to classify the plant seedlings. This classification process is often done by experts, but it is a time-consuming and labour-intensive procedure. In contrast to human examination, the automated classification of seedling methods results in a more effective way to analyze crop growth. Moreover, automating seed classification increases the precision and consistency of determining seed quality. Here, an efficient plant seedlings development classification technique is employed to improve crop production using deep learning techniques. In the initial phases, essential plant seedling images are collected from benchmark resources. The image segmentation procedure is executed in the collected images through the Transformer Unet++ (TransUnet++) technique. Once, the image segmentation procedures are completed then these images are given to the image classification phase. Here, the Residual Inception Network (ResINet) is employed for executing the plant seedling development classification procedures. Finally, the classified outcomes for plant seedling development are attained from the ResINet technique. This model is capable of learning complex features and increasing classification accuracy due to the combined advantages of residual learning and inception modules. Different experiments are executed on the developed plant seedling development classification model over existing techniques to analyze its effectiveness.