Lab 10 - nikhitasharma/RTBigDataAnalytics_Project GitHub Wiki

Task: Audio/Video Classification

Input: The video data with audio. The audio and video features from the input data are extracted and processed separately.

Output: The end results are annotated on the video input given

The Task is achieved in two different phases:

Training Phase:

Audio Classification:

In the training part the audio features are extracted from the given input data provided. The generated features are are stored to the mongoDB using REST API calls.

The stored features are retrieved by the kafka consumer to generate the Decision Model using spark Mlib.

The generated model is again stored in the mongoDB for future reference.

Video Classification:

For the given input data the video features are extracted from the given input data provided. The process of video features model generation follows the similar way of audio model generation.

Testing Phase:

Audio/Video Classification: In the testing phase the new features that are to be classified are sent to the kafka consumer from kafka producer through REST API call.

The kafka consumer parses the new audio/video features to be classified through the two different audio/video decision models generated respectively in the training phase to determine the class to which the feature belongs.

The results are annotated on the given video.