Multiscale Convolutional Neural Networks with Attention for Plant Species Recognition.

Area of Science:

• Botany
• Computer Science
• Artificial Intelligence

Background:

• Plant species recognition is vital for biodiversity conservation.
• Leaf-based recognition faces challenges due to leaf variability and scale normalization issues in current methods.

Purpose of the Study:

• To develop a novel model for accurate plant species recognition using leaf images.
• To address limitations of single-scale recognition in existing plant identification techniques.

Main Methods:

• A novel multiscale convolutional neural network with attention (AMSCNN) was designed.
• Multiscale convolution captures low and high-frequency image features.
• An attention mechanism enhances feature extraction and network training.

Main Results:

• The AMSCNN model demonstrated superior performance compared to traditional and deep learning methods.
• Extensive experiments on a plant leaf dataset validated the model's effectiveness.
• The highest accuracy achieved was 95.28%.

Conclusions:

• The AMSCNN model offers a significant advancement in automated plant species recognition.
• This approach enhances the potential for effective plant diversity monitoring and conservation efforts.