Sign in Subscribe
Artificial intelligence

Breakthrough in Mushroom Image Classification using Artificial Intelligence

Strategic Insights

— 1 min read

Researchers from the Haldia Institute of Technology have achieved significant breakthroughs in classifying mushroom images into multiple categories using deep learning-based models. The team, led by Bidesh Chakraborty, developed a novel approach that incorporates the convolutional block attention module (CBAM) with a transfer learning-based Xception architecture, resulting in superior performance compared to conventional transfer learning approaches. The model achieved validation accuracies of 95.85% and 93.68% on two different mushroom image datasets, outperforming other comparable deep learning-based systems.

Key Takeaways:

  • The research team developed a convolutional block attention-based deep neural network (CBM-DNN) for mushroom classification, which integrates the CBAM with a transfer learning-based Xception architecture.
  • The CBM-DNN achieved validation accuracies of 95.85% and 93.68% on two different mushroom image datasets, outperforming other comparable deep learning-based systems.
  • The model selectively unfreezes the last 4 blocks to enhance feature extraction, improving the model's attention to critical spatial and channel information.
  • The study highlights the potential for improved performance by experimenting with more sophisticated deep learning-based approaches and introducing more mushroom categories.
  • The research concluded that the novelty of the model lies in its unique combination of CBAM and transfer learning-based Xception architecture.

Statistics:

  • Validation accuracy: 95.85% ( dataset 1) and 93.68% (dataset 2)
  • Number of mushroom image datasets used: 2
  • Number of mushroom categories: multiple
  • Deep learning-based approaches used: convolutional neural networks (CNN), transfer learning, and convolutional block attention module (CBAM)
  • Model's generalizability: 5-fold cross-validation approach was employed to enhance the model's generalizability

Sources:

  • Scanlan, S. Convolutional block attention-based deep neural network for mushroom classification. Discover Artificial Intelligence, 2025, 5(1):1-21.
  • Haldia Institute of Technology, Department of Computer Science and Engineering (AIML)
  • Chakraborty, B., Mukherjee, R., Mandal, S. (2025). Recent Findings from Haldia Institute of Technology Highlight Research in Artificial Intelligence (Convolutional block attention-based deep neural network for mushroom classification). Journal of Engineering, 20 (2025), p 2456.
  • Springer. Discover Artificial Intelligence Journal.

Read more