Audio-based-Emotion-Detection

Sentiment Analysis using audio of a person to predict mood in real time deployed in Jupyter Web App

Requirements

1. Tensorflow
2. Keras 2.1.5
3. Librosa
4. NLTK
5. AppMode

Instructions to run

1. Open Jupyter Notebook and head over to Predict.ipynb
2. Select AppMode extension on Jupyter , to open the WebApp
3. Enter any text when prompted Ready
4. After this recording will start and Predictions will appear shortly

Installation Specifications

1. Keras version needs to be 2.1.5 as specified due to some syntax deprecations
2. Install AppMode by conda install -c conda-forge appmode
3. It is better to install every dependency using Anaconda for 0 HASSLES

The checkpoint-0.910.h5 file is uploaded to the drive link where we made our Video Submission and Presentation. That is the saved model(trained by us) which needs to be used for inference.