The flow Diagram of the system:
The system mainly consists of 2 parts: a)Real time facial emotion detection & prediction using CNN b) Recommendation of songs from the knowledge graphs
a) Real time facial emotion detection & prediction using CNN
-The model used was a CNN Sequential model for predicting emotion of user's real time facial expression input.
-The model was trained on 28,000 training images with 7 annotated emotions: sad, happy, fear, neutral, surprise, disgusted & anger.
-Later, the model was validated on a dataset of over 7,000+ greyscale annotated images.
-The Training accuracy was 72.32% and the validation accuracy was 61.34%.
-The plotss for accuracy and loss are:
b)Knowledge Grapg Representation & Song Recommendation -The knowledge graph is created on the csv file containing over 1 million spotify tracks. -The threshold value for adding the edge between the nodes can be changed accordingly and also the similarity metrics(cosine similarity can also be used) -As of now, for easier visualization and easier understanding, we have created a seperate csv file of 100 songs, named as "small.csv". The knowledge graph is created on this smaller dataset. -The predicted emotion from the CNN model is fed as input to the graph for recommending the songs.
The outputs of the system are:
Key Features:
Facial emotion recognition using CNN Knowledge graph for music and emotion representation Similarity scoring based on Euclidean distance metrics Personalized music recommendation based on user's emotional state Technologies Used:
Python OpenCV TensorFlow/Keras NumPy Pandas Knowledge graph libraries (e.g., NetworkX, RDFLib) Web camera integration