Scholarlink

With millions of workers, agriculture continues to be the most significant sector of the Indian economy. A vast variety of crops that thrive in the nation's various climates, from arid to tropical, can be grown and sold by farmers. Many farmers in India continue to rely on traditional farming practices passed down through generations, often overlooking the impact that environmental conditions can have on crop yields. Today, the soil conditions of a region don’t remain constant over the years, and a single misguided decision by the farmer may have adverse effects on the agricultural economy. This paper presents a hybrid approach that combines deep learning for soil image classification using Convolutional Neural Network and machine learning for crop recommendation using Support Vector Machine. The 32–64 CNN architecture achieves an accuracy of 96.82% while the 64–128 CNN architecture achieves an accuracy of 93% using the primary dataset.

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

DL-Driven Approach for Soil Classification and Crop Suggestion

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sakshi Sethi,
  • Manali Pusalkar,
  • Shruti Pande,
  • Ankita Lonkar,
  • Urmila Pawar

摘要

With millions of workers, agriculture continues to be the most significant sector of the Indian economy. A vast variety of crops that thrive in the nation's various climates, from arid to tropical, can be grown and sold by farmers. Many farmers in India continue to rely on traditional farming practices passed down through generations, often overlooking the impact that environmental conditions can have on crop yields. Today, the soil conditions of a region don’t remain constant over the years, and a single misguided decision by the farmer may have adverse effects on the agricultural economy. This paper presents a hybrid approach that combines deep learning for soil image classification using Convolutional Neural Network and machine learning for crop recommendation using Support Vector Machine. The 32–64 CNN architecture achieves an accuracy of 96.82% while the 64–128 CNN architecture achieves an accuracy of 93% using the primary dataset.