Emotions recognition

This project focuses on utilizing the YOLOv8 classification model to classify human emotions based on images. The overall workflow demonstrates how to efficiently train and test a deep learning model on a custom dataset, showcasing the potential of AI in recognizing subtle patterns like emotions.

Dataset Visualization:

Visualizing random samples from the dataset helped in understanding the data distribution and verifying the structure and quality of the images. This step is crucial for ensuring that the dataset is clean and well-organized, which directly impacts the training process.

Model Training:

The pre-trained yolov8n-cls.pt model was fine-tuned on the custom dataset using transfer learning. Training the model for 15 epochs with small image sizes (64x64) was computationally efficient while maintaining a balance between accuracy and speed. The use of YOLOv8 ensures modern optimization techniques and fast inference capabilities.

Prediction and Testing:

The trained model was tested on a sample image, producing class probabilities that reflect the model's confidence in its predictions. This demonstrates the model's ability to generalize and classify emotions accurately, assuming the dataset is sufficiently diverse.

Challenges and Limitations:

The dataset size and quality can significantly impact the model's performance. A larger and more diverse dataset would enhance generalization.

The small image size (64x64) reduces computational requirements but may limit the model's ability to capture fine-grained details.

Misclassification may occur in emotions with subtle differences (e.g., "neutral" vs. "sad"), which could be improved with additional training data or more complex models like yolov8m-cls.pt.

Applications:

This project has practical implications in fields such as mental health monitoring, human-computer interaction, and customer feedback analysis. By integrating this model into real-world systems, it could enable personalized responses, improve communication, and enhance user experiences.