Follow these steps to run the project locally:
-
Clone the repository:
Open your terminal and run the following command to clone the repository:
git clone https://github.com/guhyun9454/EmotionAnalyzerWebApp
-
Navigate to the project directory:
After cloning, change into the project directory:
cd [cloned directory] -
Start the application:
Run the following command in the terminal to start the application using Docker Compose:
Make sure that you installed and run Docker Desktop or Docker Daemon
docker compose up
-
Access the application:
Open your web browser and go to "http://localhost:8501" to experience the application.
-
Explore
Test with your own images or test image provided.
Accuracy may be poor for faces that do not look forward.
-
Stop the application
To stop and remove all stuffs created by
docker compose up, run the following command in the terminal:docker compose down
The system uses a microservice architecture orchestrated by Docker Compose.
- The emotion classification model is based on a CNN architecture optimized for detecting subtle facial features and expressions.
- The model was trained from the scratch on a dataset specifically designed for Korean facial features to ensure higher accuracy in the target demographic. You can find the dataset here.
This repository deploys the previously trained model as a web application. The training process involved testing multiple models and fine-tuning the architecture for optimal performance. The data was preprocessed and analyzed to ensure high-quality inputs for training. For detailed information on the model training process and to view the full report, please visit the training report repository.