Pet dog facial expression recognition based on convolutional neural network and improved whale optimization...

Area of Science:

• Computer Science
• Artificial Intelligence
• Animal Behavior

Background:

• Understanding pet dog emotions is crucial for human-animal companionship.
• Convolutional Neural Networks (CNNs) are powerful for image recognition but sensitive to parameter settings.
• Suboptimal CNN parameters can lead to slow learning and local optima.

Purpose of the Study:

• To develop an accurate and efficient method for dog facial expression recognition.
• To improve the performance of CNN models for this task by optimizing parameters.
• To investigate the effectiveness of swarm intelligence algorithms in CNN parameter tuning.

Main Methods:

• Utilized a dedicated face detector from the Dlib toolkit for facial region identification.
• Employed data augmentation to create a comprehensive dog facial expression dataset.
• Introduced random dropout layers and L2 regularization to mitigate overfitting and reduce parameters.
• Applied an Improved Whale Optimization Algorithm (IWOA) to optimize CNN hyperparameters, including dropout probability, L2 regularization parameter, and learning rate.

Main Results:

• The proposed IWOA-CNN model demonstrated superior performance in dog facial expression recognition compared to traditional methods like Support Vector Machines and LeNet-5.
• The IWOA effectively optimized the critical parameters of the CNN, leading to improved accuracy and efficiency.
• The study validated the efficacy of swarm intelligence algorithms for complex model parameter optimization.

Conclusions:

• The IWOA-CNN model offers a significant advancement in automated dog facial expression recognition.
• Optimizing CNN parameters using swarm intelligence algorithms like IWOA is a viable strategy for enhancing deep learning model performance.
• This research contributes to better human-dog communication and welfare through improved emotion recognition technology.