Abstract:
The study addresses the use of deep learning to automate the recognition of local variations
of spinach and the evaluation of their freshness, addressing key challenges facing the
agriculture industry. The study carefully preprocesses the data for model training using a
variety of datasets and models, including ResNet101, EfficientNetB1, ResNet50, VGG19,
CNN01, and CNN02. The approaches that have been proposed indicate an outstanding
capacity to recognize between different kinds of spinach and accurately determine the
condition of freshness. The ability of the models to classify spinach into different
categories according to conversion and varied evaluations of freshness, from perfect state
to minor damage or spoiling highlights the effectiveness of the models. Comparative
analyses provide information about the advantages and disadvantages of models, which
helps users identify the appropriate architectures based on specific operational
environments. The results of the study are diverse and represent many different facets of
agriculture, such as food processing quality control, an effective supply chain, and
customer satisfaction assurance. By showing the viability as well as the effectiveness of
using deep learning for crop variant identification and freshness detection in agriculture,
this research contributes to the increasing body of knowledge. Potential methods for future
research in precision agriculture include capacity, crop ability to adapt, and more general
uses. The suggested CNN01 architecture achieved a 99.34% accuracy score on the dataset,
outperforming the other models that were evaluated. For the purpose of avoiding
overfitting, the suggested algorithm is carefully trained. The accuracy, precision, recall,
and F1 score of the trained model are evaluated using a new testing dataset. The results of
the experiment show how well deep learning algorithms can be used to accurately identify
local spinach variants and detect their freshness.
