Scholarlink

This paper presents the results of the prediction of soil organic matter (SOM) content in agricultural soil samples by training and testing of various deep learning (DL) models in Python, followed by their implementation and testing on a Field Programmable Gate Arrays (FPGA). This paper investigates the transformative potential of cutting-edge technologies, with special attention to Image Processing and Image Analysis using DL models while implementing them on FPGAs to predict SOM content in soil samples. First, Deep Convolutional Neural Network (DCNN) models were trained using 180 soil sample images categorized into three classes, alongside predictor variables, with datasets partitioned into training, validation, and testing subsets. By leveraging TensorFlow and Keras Deep Learning frameworks, pre-trained weights, and data augmentation techniques, the DCNN model demonstrates exceptional performance, minimizing prediction errors. Moreover, the integration of Artificial Intelligence (AI) such as DCNN in image processing proves instrumental in precisely detecting and categorizing different soil types. This study incorporates popular DCNN models, including VGG-16, MobileNetV2, and InceptionV3, achieving notable accuracies. Notably, the proposed DCNN model surpasses the accuracy and outperforms the results reported in the literature. This showcases its efficacy in advancing SOM prediction through innovative DL approaches. Next, the integration of model optimization techniques, such as quantization and TensorFlow Lite, adds to the robustness of the findings, emphasizing the significance of these advancements in sustainable precision agriculture and environmental preservation. At last, but not least the resultant embedded FPGA-based system is not only user-friendly but also energy-efficient, representing a significant leap forward in enabling comprehensive soil analysis and fostering sustainable, precision farming practices.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

FPGA-Accelerated Deep Learning for Soil Organic Matter Prediction: An Innovative Machine Learning Approach

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Vaishali S. Patil,
  • Santosh A. Shinde,
  • Nandkishor M. Dhawale

摘要

This paper presents the results of the prediction of soil organic matter (SOM) content in agricultural soil samples by training and testing of various deep learning (DL) models in Python, followed by their implementation and testing on a Field Programmable Gate Arrays (FPGA). This paper investigates the transformative potential of cutting-edge technologies, with special attention to Image Processing and Image Analysis using DL models while implementing them on FPGAs to predict SOM content in soil samples. First, Deep Convolutional Neural Network (DCNN) models were trained using 180 soil sample images categorized into three classes, alongside predictor variables, with datasets partitioned into training, validation, and testing subsets. By leveraging TensorFlow and Keras Deep Learning frameworks, pre-trained weights, and data augmentation techniques, the DCNN model demonstrates exceptional performance, minimizing prediction errors. Moreover, the integration of Artificial Intelligence (AI) such as DCNN in image processing proves instrumental in precisely detecting and categorizing different soil types. This study incorporates popular DCNN models, including VGG-16, MobileNetV2, and InceptionV3, achieving notable accuracies. Notably, the proposed DCNN model surpasses the accuracy and outperforms the results reported in the literature. This showcases its efficacy in advancing SOM prediction through innovative DL approaches. Next, the integration of model optimization techniques, such as quantization and TensorFlow Lite, adds to the robustness of the findings, emphasizing the significance of these advancements in sustainable precision agriculture and environmental preservation. At last, but not least the resultant embedded FPGA-based system is not only user-friendly but also energy-efficient, representing a significant leap forward in enabling comprehensive soil analysis and fostering sustainable, precision farming practices.