International Journal of Computer Applications
Foundation of Computer Science (FCS), NY, USA
|
Volume 187 - Issue 15 |
Published: June 2025 |
Authors: Girish D. Chate, S.S. Bhamare |
![]() |
Girish D. Chate, S.S. Bhamare . Comparative Performance Analysis of Deep Learning Techniques for Soil Image Classification. International Journal of Computer Applications. 187, 15 (June 2025), 61-70. DOI=10.5120/ijca2025925196
@article{ 10.5120/ijca2025925196, author = { Girish D. Chate,S.S. Bhamare }, title = { Comparative Performance Analysis of Deep Learning Techniques for Soil Image Classification }, journal = { International Journal of Computer Applications }, year = { 2025 }, volume = { 187 }, number = { 15 }, pages = { 61-70 }, doi = { 10.5120/ijca2025925196 }, publisher = { Foundation of Computer Science (FCS), NY, USA } }
%0 Journal Article %D 2025 %A Girish D. Chate %A S.S. Bhamare %T Comparative Performance Analysis of Deep Learning Techniques for Soil Image Classification%T %J International Journal of Computer Applications %V 187 %N 15 %P 61-70 %R 10.5120/ijca2025925196 %I Foundation of Computer Science (FCS), NY, USA
Soil classification through image analysis has appeared as a key tool for advancing precision agriculture and land resource management. This research conducts a detailed comparative study of two prominent convolutional neural network architectures, InceptionV3 and ResNet50, applied to soil image classification. A curated dataset including diverse soil types used for model training and evaluation. Transfer learning was utilized to modify models already trained for the soil classification task, and hyperparameters were optimized to enhance performance. Used comprehensive assessment criteria, including overall accuracy, precision, recall, F1-score, and matrix of confusion analysis. The results show that InceptionV3 offers advantages in computational efficiency and faster convergence, while ResNet50 demonstrates superior classification accuracy and generalization, particularly for heterogeneous and complex soil textures. This paper provides an in-depth understanding of the trade-offs between model complexity, training dynamics, and classification performance, serving as a guideline for future applications of deep learning in soil science.